Tlre

MS-f)OS
Encyclopr,dia

Microsoft Press
'\(ashington
Redmond,
1988

Ray lluncan, General Editqr

i Foreword by Bill Gates

I
Foreword
more
Microsoft's MS-DoS is the most popular piece of software in the world' It runs on
worldwide and is the foundation for at least 20'000
than 10 million personal computers
applications the largest set of applications in any computer environment As an industry
siandard fo.
-
the famif of8086-based microcomputers' MS-DOS has had a central role in
the personal computer revolution and is the most siSniFicant and enduring factor
in fur-
thering Microsofti original vision a computer for every desktop and in every home The
-
challe'nge of maintaini-n$ a single operating system over the entire range of8086-based
to meet-
microco'mputers and applications is increctible, but Microsoft has been committed
ing this ch;llenge since the release of MS-DoS in 1981 The ttue measure of our success
inihis effort is MS-DOS's continued prominence in the microcomputer industry
have entered the
Since MS-DOS'S creation, more powerful and much-improved computers
marketplace, yet each new version of MS DOS reestablishes its position as the foundation
for new applications as well as for old. To explain this extraordinary prominence' we must
look to the origins of the personal computer industry The three most significant factors in
the creation oiMS-DOS were the compatibility revolution, the development of Microsoft
and IBM's deci-
BASIC ancl its widespread acceptance by the personal computer industry,
sion to build a computer that incorporated 16-bit technology'

The compatibility revolution began with the Intel 8080 microprocessor' This technolog-
ical breakthrough brought unprecedented opportunities in the emerging
microcomputer
in power, sPeed, and cost ofdesktop com-
industry, promising continued improvements
puting. In the miniiomputer market, every hardware manufacturer had its own special
instruction set ancl operating sfstem, so software developed for a specific machine was in-
compatible with the machines of other hardware vendors This specialization also meant
tremendous duplication of effort each hardware vendor had to write language compilers'
- tools to fit its particular machine Microcomputers
databases, and other develoPment
based on the 8080 microprocessor promised to change all this because different manu
facturers would buy the same chip with the same instruction set'

Frorn 1975 to 1981 (the 8-bit era of microcomputing' Microsoft convinced

virtually
Radio Shack, Commodore, Apple' and dozens
every personal computer manufacturer
-
to build Microsoft BASIC into its machines For the first time' one common lan
of otirärs
-
g.r"g. .,.tt u..o"" dl han]ware venclor lines. The success of our BASIC demonstrated the
To their great benefit, users were finally able to move appli-
iJ,,i*"g., or.o-patibility:
cations from one vendor's machine to another'
disk drive
Most machines produced during this early period did not have a builFin
and later fixed clisks' became less expensive and more
C.adually, howä.,er, floppy disks,
including \fordStar and dBASE' entered
.o--on, "
.tlmUei oi disk-Uasea programs,
".rd that could accommodate these develop-
the market. A standard disk operating system

-.*" b.."-" ."tremely imptrtant, Lading Lifeboat, Microsoft, and Digital Research all to
support CP/M-8O, Digital Research's 8080 DOS'

Foreuorcl xüi
 The S-bit era proved the importance of having a multiple_manufacturer standard that
permittedthe free interchange of programs. It was important that
software designed for
the new 16-bit machines have this same aclvantage. N; personal
computer manufacturer in
1980 could have predicted with any accuracy how quickJy
a third_pany software industry
would grow and get behind a strong standarci-a slancrarcr that wourd
be the software
industry's lifeblood. The intdcacies of how MS-Dos became rhe mosr
common 16-bir
operating system, in part through rhe work we did for IBM, is not the key point
here. The
key point is that it was inevitable for a popuJ:rr operaring sy.stem to
cmerge tbr the 16_bit
machine, just as Microsoft's BASIC had prevailecl on the 8_tit systems.

It was overwhelminpJly evident that the personal computer had reachecl

broad acceptance
in tlre market when Time in 1982 named the personaicomputer ,,Man
of the year.,,MS
DOS was integral to this acceptance and popularity, and we have
continuecl to adapt
MS DOS to support more powerful computers without sacdficing
the compatibility that is
essential to keeping it an industry standard. The presence ofthe"SOjA6
microp.o.."ro.
guarantees that continued investments in Intel_architecture
software will be worthwhile.
Our goal with The MS-DOS Enclctopedra is to provi<1e the mosr
thorough and accessible
resource available anywhcre for MS_DOS programmers. The length
ofthis book is many
times greater than the source Iisting of the first version of MS-Dö-evidence
of the
growing complexiry and sophisrication ofthe operating system.
The encyclopedia will be
especially useful to software developers faced with preieiving continuity
yet enhancing
lhr pofldbililv ol I heir rpplir ar ions
Our thriving industry is committed to exploiting the advantages
offered by thc protected
mode introduced with the 80286 microprocessor and the virtuar mode
introclucecr with the
80386 microprocessor MS DOS will continue to play an integral
part in this effort. Faster
and more powerful machines running Microsoft OS,/2 meanän
exciting tuture of multi
tasking systems, networking, improved levels ofdata protection,
bette; hardware memory
management for multiple applications, stunning graphics systems that
can display an inno_
vative graphical user interface, and communication subsystems. MS
DOS version 3, which
runs in real mode on 902g6-based and 803g6-based machines, is
a vital rink in the F-am y
API of OS,/2. Users will continue ro benefit fiom our commitment to
improved operating_
system performance and usability as the future unfolds.

Bill Gates

The.US DOS Erutcta;*rha

Preface
operating
In the space of six years, MS-DOS has become the most widely used computer
system in the worlä, running on more than 10 million machines lt has Srown' matured'
system that can support networking'
a'nd stabilized into a flexible, easily extendable
peripheral device,.and even CD ROMs containing
graphical user interlaces, ncarly any
will be with us for many years to come
äassive amounts of online information. MS-DOS
low-cost, 8086/8088 based machines
as the platform fo. applications that run on
into its
Not surprisingly, the success of MS-DOS has drawn many writers and Publishers
and applications
orbit. The nulter ofbooks on MS-DOS ancl its commands, lang,ages,
yet another book on
dwarfs the list oftitles for any other operating system Why' then'
MS-DOS? And what can we say aboutihe operating system that has not
been said already?
in mind:
First, we have written and eclitecl The MS-DOS Encyclope'Jia w\th
orLe
^Lrdience
we have therefore been free to bypass elemen-
the community of working programmers'
tary sublects such the number ofbits in a byte and the interpretation of hexadecimal
as
ex
,r.,-b"r.. Inrt."d, *. have emphasized cletailed technical explanations, working code
into new applications' and a systems view of
amples that can be adaptecl anä incorporated
even the most common MS-DOS commands and utilities'

seconcl, because we were not subiect to size restrictions,

we have explored topics in depth
and error handling'
that other MS-DOS books mention only briefly, such as exception
and install-
interupt driven communications, debugging strategies, memory management'
able device drivers. Ve have commissioned definitive articles
on the relocatable object
translators, the operation ofthe Microsoft Ob-
-oa.,f., g".t..rr.O by Microsoft language utilities-Ve have even interviewed the key
ancl-stay iesilent
1.c, rintä., ,na,.tminate
äweloperi of trls,oos and drawn on their files and bulletin boards to offer an entertain-
operatin€l system'
ing, illustrated account ofthe oriSins of Microsoft's standard setting
programmers and
Finally, by combining the viewpoints and experience of non Microsoft
and resources of Microsolt software developers' and the publishing
*riterr, th.
""pertise unique and comprehensive reference
krow-how of Microsoft Press, we have assembled a
utilities ln many instances' the manu-
to MS-I)OS services, commands, directives, and
,.rip," tto,r. U"." .eviewed by the authors ofthe Microsoft tools described'
'!!e have made every effort during the creation ofthis book to ensure that its contents are
and omis-
it*.iv -J,t""*"rihy. L, " -o,i nf this size, however' it is inevitable that errors so that
plcase bdng them to our attention
sions will occur' Ifyou cliscover any such errors,
they can be repaired in future priniings and thus aid your fellow programmers To this
has cstablished a bulletin board on MCI Mail for posting
corections
.r,i, lli.rorufipt.ts
and comments. Please refer to page 'üzll for more information'

Ray Duncan
The Development of MS-DOS

To many people who use personal computers, MS,DOS is the key that unlocks the power
ofthe machine. It is thcir most visible connection to the hardware hidden inside the
cabinet, and it is through MS-DOS that they can run applications and manage disks and
disk files.

In the sense that it opens the door to doing work with a personal computer, MS-DOS is
indeed a key, and the lock it fits is rhe lntel 8086 family of microprocessors. MS DOS and
the chips it works with are, in fact, closely connected so closely that the story of
MS DOS is really pafi of a larger history that encompasses not only an operating system
but also a microprocessor and, in retrospect, part ofthc explosive growth of personal
computing itseli
Chronologically, the hisrory of MS-DOS can be divicled into rhree parts. Firsr came rhe
formation ofMicrosoft and the events preceding Microsoft's decision to develop an
operating system. Then came the creation olthe first version of MS-DOS. Finally, there is
the continuing evolution of MS-DOS since its release in 1981.

Much ofthe story is based on technical developments, but dates and facts alone do not
provide an adequate look at the past. Many people have been involved in creating MS DOS
and directing the lincs alon€l which it continucs to grow. To the extent that personal opin-
ions and memories are appropriate, they are included here to provide a fuller picturc of
the origin ancl clevelopment of MS-DOS.

Before MS-DOS
The role of Intcrnational Rusiness Machines Corporation in Micr()s()ft's decjsion to create
MS-DOS has been well publicizcd. But events, like invcntions, always build on prior ac-
cornplishments, and in this respect the roots of MS-DOS reach farther back, to four iurd
warc and sofiware developments ofthe 1970s: Microsoit's clisk basecl and stand akrne
versions of BASIC, Digital Research's CP/M-80 operating system, the emergence ofthe
i3086 chip, ancl a disk opcrating systcm for the 8086 developed by Tim Parerson at a hard,
ware company called Seattle Computer Prcxlucts.

Microsoft and BASIC

On the surfäce, BASIC and MS-DOS might seem to have littlc in common, but in tcrms of
filc managcment, MS-DOS is a direct descendant of a Microsoft version of BASIC called
Stand alone Disk IIASIC.
Before Microsoft even bccame a company, its founders, Paul Allen and Bill Gates, dc,
velopecl a version of BASIC for a revolutionary small computer named the Ältair, which
was introduced inJ.rnuary 197i by Micro Insttimentation Telemetry Systems (MI't'S) of

Sectian I The Deüelapment oJ Ms-DOs

1975

The Altaif. C.hristetBd one euefüng shortlf belore its alü)earance ()n
the .opr o/ popular Elecrronics
m.agazite, the computel aas named
for the t ight,s clestination ofthe stars&rp enterirtse. The phorogfaph
clearlj shobs the input srDitches ofi thefront
|lanet of the cabinei .

Mexico. Though it has long been ectipsecl by otheq more powerfut

1ll:Oi:rl*. I.:
makes and mrrdels. the Altair w;s the first ,,personal', computer to appear in
an envir()n
ment dominated by minicomputers ancl mainframes. tt was,
simply, J metal box with a
panel of swirches and Iighr5 for jnpur and ourpur, a power
suppli, ä'motnerboard with 1g
srors, anct rwo lloards. One board was the central processin€a
unit, with the g_bit Intel gog0
microprocess<tr at its heart; the other boartl
f,rr_,vidcd 2io bytes ofrandom_access memory.
This miniature computer had no kcyboarcl, no moniror,
nnj ,,o a.uil. fo. permanent
storage, but it did possess one great advantage: a pdce
tag of $392
N'ow, given the hindsight of a little more than a decade
of microcomputing history, it is
easy to see that the Ahai/s combination.of smali size and affordability was the thin edge
of a wedge.that, in just a few years, would move everyday computing'power
away from
impersonal monoliths in climate-controlled rooms and onto
the clesis of millions of
people. In 1975, however, the computing environmcnt
was still primarily a matter of data
processing fbr specialists rather than personal computin[l
for everyone. Thus when 4 KB

The llS DOS E,tc.lctope.tia

 I
Intet's 4OO4, 8008, and 8O8O chips. At the top left is the bit 4004, ahich üas named.fof the approximate
I
number of otcl-fashioned transistors it replaced. At the bortom left is the bit aooa, uhich addfessed 16 RB of
memory); this ulis the chip used in the Traf-O-Data tape-reader buih bl Pa l Gilbert At the riSht is üe 8080,
afaster 8-bit chip that could address 64 KB o.f memori The brain ofthe MITS Ahaia the 8080 orts, in ntan|
respects, the chip on uhich the personal computing industry oas builr. The 4O(M an l SOAS chips üere
deLEloped earll in the 1976; the 8080 appeared in 1974 .

merrory expansion boards became av:rilable for the Altait the softwarc needecl mosr by its
llsers was not a $()rcl processor or a spreädsheel, but a programming language and the
-
language first clevclopecl for it was a vcrsion of BASIC written by Bill Gates ancl Paul Allen.

Gates anciAllen tr:rcl become friencls in their teens, whilc .rttending Lakesidc School in
Seattle. They shared an intense intcrest in computcrs, ancl by rhe time Geles was in the
tenth gracle, they and another liiencl namecl I'aul Gilbert l.tad fbrmed a company callcd
Traf O-Data to proclucc a mrchine that automaled the reacling ol16-channel, 'l-digit.
binary-coded clecimal (l]CD) tapes generated by traffic lnonitoring re'orders. This ma
chine, built by Gilbert, was basccl on the Intel 8008 microprocessor, the predeccssor
ofthe B0B0 in the Altair.

\e.1i|n I Tht D?t,l.l,nLntoJ,ll' DOS

7975

TheJanuaty 1975 co,er af poplrlat

HO\/ I1O IREAD' FM TUXER Ele.iofücs magazine, -featuring the
SPECIFICATIONS
machine that caught the imaginatians

hpular
ffi
Electrtrnics af thousands af like minded electron-
ics enlhusiasts
- among them, paul
Allen and llill Catp:

PNO'ECT BREAKTHNOUGH !

World's First Minicomputer Klt

to Rhal Commerclal Models...
ATTAIR 8800" savE OVER $rooo

atso I x rHts tssuE:

.An Under-$90 Sct.n rlc Cllcuhior plol"ct
I a. CCD's-TVClncra lube iucccrror?
Thlllttor-Controlled phciolte;hcrr
IEST REPORTST
Technics 2O0 Speaker System
Pioneer RT.lotl 0pen-Reet Recoder
Tram Diamond-4o CB AM Transceiver
Edmund Scientific "Kirtian. photo Kit
Hewlett-Fackard 53Bl FrEquency Counter

Although it was too limitcd to serve as the central processor for a general_purpose
compu_
ter, the 8008 was undeniably the ancestor of the g0g0 as lar as
its ähitecture ancl insrrrrc_
tion set were concerned. Thus Traf-O-Data,s work with the g00g gave
Gates ancl Allcn a
head start when they larer developecl lheir version of BASIC for
th"e elralr
Paul Allen learned of the Altair from the cover story in theJanuary
1975 issue of popular
Electronics magazine. Ailen, then an employee ofHoneywell in äoston, convinccd
Gates,
a student at Harvard University, to develop a BASIC for the new computer The
two wrote
their version of BASIC for the 8080 in six wceks, and Allen flew to New Mexico to .lemon_
strate the language for MITS. The developers gave themselves the
company name of
Microsoft and licensed thcir BASIC to MITS as Microsofr,s fircr product.
Though nor a direct forerunner of MS DOS, Aftair BASIC, like the machine for
which it was
developed, was a landmark product in the history ofpcrsonal conputing.
On an()ther
level, Altair BASIC was also the first link in a chain that led, somewhat
circuitously, to Tim
läterson and the disk opcrating systern he developed for Seattle Computer products
fbr
the 8086 chip.

The MS-DOS Enqclopedia

 1976

cclqrllR rtrEs/&!v, !975

c@ C<'aaArl fr, "*- *" -,, h,t

h61t/)-
[.r,,r..) \
l,-".f ,* ,d lzt1l-') ] that yoü k.y in na.u:lly (I.33 th.n
25 byt.r). rhis
I.v,.dve.a 5&.,1-,. l]
Ltn-\ !E r{d.! (rh. 'ibi'! rdd..)
rrrrb.$dtI20byt.'.
rhl.h

\,".
LI
I ro' .udl. d3!.rt. lo.diq !h.
[email protected] .nd ch.ck3ur !ö.d.r! tirl
of this rilr b. .r-
'Rr,,1 a 6ae -+- -J \ I a> pr.lr.d ln d.t!l! i! . cov.. 9.ck.3"
z<." th.t rll! 3o oui flrh arr softf.r..
f.\ ).,)
I t-^"] !'^rr" "-.,atl.:.6ajtt y *',".11<
7 b'11e, q t +Lp
lo.la.drng..n.ch..k3w.a
it . :ho.t prcsrd I

1 u1k' 4-z +L- ^*- vol"a (? byt.3-l! loi by!. of

.Pzp+-l +* e"Ä ß.,J/a -
,2 hiah brt. of
f+eqr+a) -t,, -t r"''"
t"-,
Lt' t ctl8tr ädd..33)
laE1t l

I 2 br it l."rlq.\
lv't-t - R^L ;g.t input rc.dy bit
P-
".^l! +". ."J -,'h
frra e"{ r,-'i1 1,""1,^ +
l.dz r|d<a (n'f"dt - t.e)
'\t
f.rxrorl ftoat ft.z!+ taal Q"it-r.
t
sidc\- (?? byr.!)
[FsErbE b"1th "t s+-.r i'1,.e-t t".o.,+' +":"
f r
tunch
0q3
. p.p!r r.p. rlth r..,1.r,
3t. byt.'
th. byr! to b.
13e-'at; .- k-{ s+,,.q ",-e2 sfr3lr i:r!i3 sur.
sl?iv G5 J ' lodd". is in i3
su.. yöu doi t eip.
fr^eusrz) t^,1u":1 onozln,- fa . -1ia.. by loadin. on toe or it.
ro .un thls as.in ch.n.. CHCIOC

'\^\ S-t olR o tt^,' l- <'f-e

^.,1,- t^a.'4-?",1-.'. O 'O o>
, t"- .r.' l. L .. tr 8i)
/".+v-*.'o

On the le.ft, Bitl Gates's orisinal handturitten nales describin| memorv contis
rationfar Allait BASIC On
thefight,ashoflboatstrapprasramafittenbfGdtes'forAttairuse|s;pubtishe.tintheJul!1975e.1itionofthe
MITS üser neusletter, Computer Notes.

From paper tape to disk

Gates and Allen's early BASIC for the Altair was loaded fiom
paper tape after the boorsl'ap
to load the tape was entered into memory by flipping switches on
thc front panel of the
for the-
computer In late 1975, howevet MITS decided to release a floppy disk system
1976
Altair-the first retail floppy-disk system on the market As a result' in Februaty
version of
Allen, by then Director oisoitware ior MITS, asked Gates ro wrife a disk-based
files'
AltairBASIC. The Altair had no operating system and hence no mcthod of managing
It would, in
so the disk BASIC would have to include some file-management loutines'
effect, have to function as a rudimentary operating system

Sectian I: The DeueloPment af MS-DOS

 1977 197A

Microsoft, 1 978, Albuquerqte,

Nea Mexico. Top rou, left to right:
SteLe Wood, Bob Wallace, Jt n Lane.
Middle ro@, W to right: Bob O'Rea/,
Bob Greenberg, Marc McDonatd,
Gordon Letuin. Bottom rou, left to
right: Bill Gates, Andrea leuis,
Marla Wood, paul All"n .

cares, still at Ha*ard Itniversity, agreed to write rhis vercion

of BASIC for MITS. He went
to Albuquerque and, as has often been recounted, checked
into the Hilton Hotel with a
stack ofyellow regal pads. Five days rarer he emerged,
yetow pads filled with the code for
the new version of BASI.. turiving ar MITS with tÄe code anda --
request to be left alone,
cates began typing and debugging and, afrer anorher five days,
haä oisk Basrc running
on rhe Altair

This disk-based BASIC marked Microsofr's entry inro rhe business of languages
for per_
sonal computers-not oniy for the MITS Altair, but also for
D"tu
Terminals Corporation and General Electric. Along the way, ",r.h.o^p".ri..-u"
Microsoft BASIC took on
added features, such as enhanced mathematics cafabilities, and,
more to the point in
terms of MS,DOS, evolved into Stand alone Disk BASIC, procluced
for NCR in 192.
Designed and coded by Marc McDonald, stand arone Disk BASIC,
incruded a file-
management scheme called the FAT, or file allocation table that used
a linked list for man-
aging disk files. The FAT, born during one ofa series ofdiscussions
between McDonald
and Bill Gates, enabled disk-allocation information to be kept
in one location, with
'thained" references pointing to the actual storage locations on disk.
Fast anj flexibte,
this file management strategy was later used in a stand-alone version
of BASIC for the g0g6
chrp andeventually, through an operating system named M_DOS, became
the basis for the
file-handling routines in MS-DOS.
M-DOS
During 1977 and 1978, Microsofr adapted borh BASIC and Microsoft FoRTRAN
for an
increasingly popular 8 bit operating sysrem called Cp/M. At the end
of197g, cares and
Allen moved Microsoft from Albuquerqr-re to Belleme, Vashingon. The company
con_
tinued to concentrate on programming languages, producing versions
of BASIC for the
6502 and the T19900.

The MS DOS Enc),clopedia

 1978

A Microsoft aduertisement from the

January 1 979 issue of Byle magazine
fie tioning same Producß and the
machines lhey ran an. In the lauer
righl corner is an announcement oJ
lhe compafiy s moue to Belleuue,

Onrv @ @rr@oiy eh he Poe w[h

!.firc€ ld nlc@t@ss.

Vr,l€\., .i 1|ta Fia-:ra .'

Cal{) t . lret :. 'IJ .i
c,üimi r€' :irl.r-t . 5e r

N4BASIC - NFW RIiEA!! ]E

ANS1 74COEOta0 s.'N

1:4 t.!
!.crryDst 3.P od

During this same period, Marc McDonald also worked on developing an 8-bit opcrating
system called M-DOS (usually pronounced "Midas" or "My DOS '). Although it never
became a real part of the Microsoft product line, M-DOS was a true multitasking operating
system modeled after the DEC TOPS-10 operating system M DOS provided good perfor-
mance and, with a more flexible FAT than rhat built into BASIC, had a better fileJtandling
structure than the up-and-coming CP/M operating system At about 30 KB, howevet
M-DOS was unfortr,rnately too big for an 8-bit environment and so ended uP treing rele-
gatedtothebackroom.AsAllendescribesit"'Tryingtodoalarge,fullblownoperating
;ystem on the 8080 was a lot of work, and it took a lot of memory, The 8080 addresses only
64 K. so wirh the success of CPIM, we finally concluded that it was best not to press on
with that."
CP/M
1976 through 1978, both users and developers of per
In the volatile microcomputer era of
sonal computers quickly came to recognize the limitations of running applications on top
of Microsort,s stand-alone Disk BASIC or any other language. MITS, for examplc, scheduled

Sectian I: The DercloPment af MS-DOS

 1974

aJuly 1976 release date for an independent operating system for its machine
that used the
code from the Aftair's Disk BASIC. In rhe same yerr, bigital Research,
headed by Gary
Kiidall, released its Control program/Monitor, or CplM.
cPlM was a typical microcomputer software product ofthe 1970s in that
it was n ritten bv
one person, not a group, in response to a specific need that had not yer heen
fiucJ. One of
the most interesting aspects of CplM's history is that the software was
developed several
years before its release date several years befbre the harclware on which it
would be a standard became-actually,
commercially available.
In 1973' Kildatl, a professor ofcomputer science at the Naval postgraduate school
in
Monterey, California, was working with an g0g0_based snrall computer given
him by Intel
Corporation in return for some programming he had done for the comjany Kildall,s
machine, equipped with a monitor and paper-tape reader, was certainly
advanced for the
time, but Kildall became convinced that magneticdisk storage woulcl
make the machine
rven morc efficirnt thJn it q d5
Trading some programming for a disk <1rive from Shugart, Kildall fißr
attempted ro builcl
a drive controller on his own. Iacking the necessary engineering
ability, he contacte.l a
friend, John Torode, who agreed to handle rhc hardware aspecrs of
interfacing the compu_
ter and the disk drive while Kildall rvorked on the softwareportion
the refinement cf an
operating system he had written earlier tllat year. The result $,,as Cp,/M.
The version of CP/M developed by Kildall in 1973 rrnderwent several reflnements.
Kilclall
enhanced the CP/M debugger and assembler, adcled a BASIC interpreter,
and clid some
work on an editor, eventually developing the procluct that, from aÄour 1977
until the ap_
pearance of the IBM Personal Computer, set the standaral for g_bit
microcompurer opcrat-
ing systems.

Digital Research's Cp/M included a command intelprcter called CCp (Console

Commancl
Processor), which acted as the interfäce berween the uscr ancl
the operatrng system itself,
and an operations handler called BDOS (Basic Disk Operating Systen.r),
r,hich was
responsible for file storage, directory maintenance, and other .such housekeeping
chores.
For actual input and output-disk L/O, screen display, print requests,
ancl so on _Cp/M
included a BIOS (Basic Input/Output System) tailorecl to the requirements
ofthe hardware
on which the operating system ran.

For file storage, CP/M used a system of eight-sector allocation units.

For any givcn file, thc
allocation units were listed in a directory entry that included the filename and a table giv
ing the disk locations of 16 allocation units. If a long file required morc than 16
alrocation
units, cP/M created additionar directory entries as required. S'ra files courct be accessed
rapidiy under this system, but large files with more than a single directory entry could re_
quire numerous relatively timc-consuming disk reacls to find needed informarion.

At the time, howevet CP/M was highly regarded ancl gainecl the support ofa
broad base of
hardware and software developers alike. euitc poweiful for its size
iabout 4KB), ir was, in
all respects, the undisputed standard in the g-bit world, and remainecl so until. ancl even
after. the appqsra6(e ol lhe c08o.

10 The,US DoS Enqtctapedia

 1974

The 16 bit Intel 8086 chip, introdAced in 1978.

MuchJaster andfar more pounrful than its g-bit
predecesmr the 8O8O, the 8086 had the ability to
I
I
address one rnegabJte o.f mernorJ) .

!,

The 8O86
Vhen Intel released the 8-bit 8080 chip in 1974, the Altair was still a year in the future.
The 8080 was designed not to make computing a part oleveryday life but to make house-
hold appliances and industrial machines more intelligent By 1978, when Intel introduced
the 16-bit 8086, the microcomputer was a reality and the new chip represented a majof
step ahead in performance and memory capacity. The 8086's full 16-bit buses made it fast-
er than the 8080, and its ability to address one megabyte ofrandom-access memoty was a
gianr srep bcyond the 8080's 64 KB iimit. Although the 8086 was not compatible with the
8080, it was architectumlly similar to its predecessor and 8080 source codc could be me
chanically translatecl to run on it. This translation capability, in fact, was a najor influence
on the design ofTim Paterson's operating system for the 8086 and, through Paterson's
work. on the first released version of MS-DOS.
When the 8086 arrived on the scene, Microsoft, like other developers, was conlionted with
two choices: continue working in the familiar 8 bit world or turn to the broader horizons
oflerecl by the new 16 bit technology. For a time, Microsoft did both Acting on Paul Allen's
suggestion, the company developed the SoftCard for thc popular Apple II, wltich was
based on the 8-bit 6502 microprocessor. The SoftCard included a 280 microprocessor and
a copy of CP/M-80 licensed from Dilaital Research !?ith the SoftCard, Apple II users could
run any program or language designed to run on a CP/M machine.
It was 16-bit technology, however, that held the most interest for Gates and Allen, who
believed that this would soon become the standard for microcomputers Their optimism
was not universal-more than one voice in the trade press warned that industry invest
ment in 8 bit equipment and software was too great t() successfully introduce a new slan-
Microsoft, however, disregarded these forecasts and entered the 16-bit arena as it
darc.l.
hacl with the Altair: by developing a stand-alone version of BASIC for the 8086

Section I: The Deuelopment ,|MS'DOS 11

 1979-1980

At the same time and, coincidentally, a few miles south in Tukwila,

\X/ashington, a major
contribution to MS DOS was taking place. Tim patemon, working
at Seattle"Computer
Products, a company that built memory boards, was developing
an g0g6 CpU a"rd fo, u".
in an S 100 bus machine.
86-DOS
Parerson was introduced to the g0g6 chip at a seminar held
by Intel in June 197g. He had
attended the seminar at the suggestion ofhis employer, Rod Brock
of ieattle Computer
Products. The new chip sparked his interest because, as he recalls, ,,alr
its instructlns
worked on both 8 and 16 bits, and you didnt have to do everything rhrough
rhe accumu_
lator. It was also real fasr- it could do a 16_bit ADD in three clocks.,,

After the seminar, paterson_again with Brock,s support_began

work with the g086.
He finished the design of his firsr 8086 CpU board inJ-an uary
I9j9 andby late spring had
developed a working CpU, as well as an assembler and an gb86
monitor. InJu";; p^:;;;"
took his system to Microsoft to try it with Stand alone BASIC, and
soon after, Microsoft
BASIC was running on Seattle Computer,s new board.

During this period, paterson also.received a call from Digital Research

asking whether
they could borrow the new board for deveroping cp/M-gz.
Though seattle computer did
not have a board to loan, paterson asked when Cp/V_g6 would
be reacly. Digitalls represen-
tative said December 1979, which meant, accordin[a to pate.son,s
dinry, ,,-e;ll have to live
wirh stand-alone BASIC ror a few months after weitart shipping the
cpt, but thcn we'' be
able to switch to a real operatinfa system.,,

Early inJune, Microsoft and Tim pate^on attendecl the National

computer conference
in New York. Microsoft had been invited to share Lifeboat Associates,
ten_by_ten foor
boorh, and paterson had been invired by paul Allen to show
AAVC -"" ing o., ,., S" fOO
8086 system. At that meeting, patemon was introduced to
Microsofi,s M Dos, which he
found interesting because it used a system for keeping track oldisk
files_the FAT devel
oped for Stand-alone BASIC-rhat was different fiom anything
he had encountered.
After this meeting, parerson conrinuecl working on the g086 board, and
by the end of the
yeat Seattle Computer producrs began shipping the CpU v..ith a BASIC option.
\rhen CP/M 86 had still not become available by April 19g0, Seanle Computer producrs
decided to develop a 16-bit operating system of its own. Originally,
three operaring sys_
tems were planned: a single user system, a multiuser version, anda
small int".i_ iroiu.t
soon inforrnally christened eDOS (for euick and Dirry Operating Sysrem)
by paterson.
Both Paterson (working on eDOS) and Rod Brock knew that a standard
operating system
for the 8086 was mandatory if users were to be assured of a wide range
ofapplicalon soft_
ware and languages. cplM had become the standarcl for g-bit machines.
so the abilitv ro
mechanically translate exisring cplM apprications to run on a 16 bit sysiem
be.n-. ärr. of
Paterson's major goals for the new operating system. To achieve
this compatibility, the sys-
tem he developed mimicked Cp/M-80,s functions and command structure.
incluiing its'
use of file control blocks (FCBs) and its approach to executable files.

12 TheMS-DOSEntytopedia
 1980

An aduertisement for
the Seattle Computer
GO 15.BIT NOW-WE HAVE MADE IT EASY
Products 8086 cPU,
86-DOS; published

8
8086
Mhz.2-cerd CPU Set
uith
in the December l98O
issue af P,yte .

wnHa6.Dos@$595
ASSEMELEO ESTED GUAiANIEEO

cdqbnedroraffe or daycck) apa

8/16 16.8IT MEMORY

l3 *': f::i ":'i".::'"1::';. "'

A

At the same time, however, Paterson was dissatisfied with certain elements of CP/M' one
of them being its file-allocation system, which he considered inefficienl in the use of disk
space and too slow in operation. So for fast, efficient file handling, he rrse<l a File allocation
table, as Microsoft had done with Stand-alone Disk BASIC and M-DOS fle also wrote a
translator to translate 8080 code to 8086 code, and he then wrote an assembler in 280
assembly language and used the translator to translate it.

Four months alter beginning work, Paterson had a functioning 6 KB operating syslem,
officially renamed 86-DoS, and in September 1980 he contacted Microsoft again, this time
to ask the company to wdte a version ofBASIC to run on his system

Section I: The DeuelLtlment of MS-DOS 13

 i980

IßM
'while Paterson
was developing 86-Dos, the third major element leacling to the creation of
MS-DOS was gaining force at the opposite end ofthe country. IBM, until rhen seerningly
oblivious to most ofthe developments in the microcomputer world, had turned its atten_
tion to the possibility of developing a low-end workstation for a market it knew well: busi_
ness and business people.

On August 21, 1980, a study group ofIBM representatives from Boca Raton, Florida, visited
Microsoft. This group, headed by a man namedJack Sams, told Microsoft of IBM,s inreresr
in developing a computer based on a microprocessor. IBM was, however unsure of micro_
computing technology and the microcomputing market. Traditionally, IBM relied on long
development cycles typically four or five years and was aware that such lengthy
- fit the rapidly
design periods did not -
evolving microcomputer environment.
One of IBM's solutions-the one outlined by Sams's group to base the new
-was hardware was available,
machine on products from other manufacturers. All the necessary
but the same could not be said of the software. Hence the visit to Microsoft with the ques_
tion: Given the specifications for an 8-bit computer, could Microsoft write a ROM BASIC for
it by the following April?

Microsoft responded positively, but added questions of its own: $trhy introduce an g_bit
computer? \ü/hy nor release a 16 bir machine based on Intel,s 9086 chip instead? Ar rhe end
of this meeting-the first of many-Sams and his group returned to Boca Raton with a
proposal for the development ofa low-end, 16-bit business workstation. The venture was
named Proiect Chess.

One month later, Sams returned to Microsoft asking whether Gates and Allen could, stili
by April 1981, provide not only BASIC but also FORTRAN, pascal, and COBOL for the new
computer This time the answer was no because, though Microsoft,s BASIC had been
designed to run as a stand-alone product, it was unique in that respect_the other lan_
guages would need an operating system. Gates suggested Cp/M_g6, which was then still
under development at Digital Research, and in fact made the initial contact for IBM. Digital
Research and IBM did not come to any agreement, however.

Microsoft, meanwhile, still wanted ro write all the languages fo. IBM_ approximately 400
KB ofcode. But to do this within the allotted six-monrh schedule, the company needed
some assurances about the operating system IBM was going to use. Further, it neecled
specific information on the internals ofrhe operating bystem, because the ROM BASIC
wouid interacl inrimareLy wrlh rhe BIOS.
The turning point
That state of indecision, then, was Microsoft's situation on Sunday, September 2g, 19g0,
when Bill Gares, Paul Allen, and Kay Nishi, a Microsoft vice president and president of
ASCII Corporation inJapan, sat in Gares,s eighth-floor corner office in the Old National
Bank Building in Belle*re, 'Washington. Gates recalls, ,,Kay and I were just sitting rhere at
night and Paul was on rhe couch. Kay said, ,Got to do it, gor ro do it., Ir was only i0 more K

74 The MS DOS Encyclopedi.l

 1980

40O It wasn't that bi8

oicode at most- actually, it turned out to be 12 more K on top ofthe
\üe'd always wanled to do a low-end oPemting
a deal, ancl once Kay said it, it was obvious
going to do
,yr,ä, -. f."a ,p"cs for los' end operating systems, and we knew we were
one up on 16-bit."
proposal to IBM' Their
At that point, Gates ancl Allen began looking again at Microsoft's
a linker To add an
estimaäd 400 KB ofcode included four languages, an assembler' and
already knew ofa
operating system woulcl require only another 20 KB or.so, and they
Nishi
J."rt,"gäaa f". the 8086: Tim Pa;erson's 86-DoS The more Gates' Allen' and the more
computer'
,"tt"a tir, nignt ofrout developing an oPerating system for IBM's new
possible even preferable idea became
- -the
to tell him that
Allen's first step was to contact Rocl Brock at Seattle Computer Produ'ts
operating system and that the company had
Microsoit wanted to clevelop and market ScP's
Products, which was not in the business of
an OEM customer for it. Seattle Computer
to Microsoft Eventually' SCP sold the
-"rt oirlg ,of,*"re, agreed and licensed 86-DoS language licenses' and a license back
op.rrrir'tg".y*.- ,o Microsoft for $50,000, favorable
from Microsoft to use 86 DOS on its own machines
Ifl October 1980. with 86 DOS in hand, Microsoft submitted another ploposal to IBM. This
the new computer'
time the plan included both an operating system and the languages for
the operating system wele
Time was short and the boundaries between the languages and
ofthe operating
.,.r.t""r. .n ttll..o"oft explained that it neecled to control the development
signed the
in nra., u, guarantee cielivery by spring of 1p81 In November' IBM
"vri.-
contract.

CreatingMS-DOS
and Bill Gates' Paul
At Thanksgiving, a prototype of the IBM machine arrived at Microsoft
af f.", sob ö'near began a scheclule of long, sometimes hectic days ancl
""AlptiÄ".ity,
total immersion in the project. As O'Rear recalls, "If I was awake'
I was thinking about

the project."
on the new machine This was
The first task handlecl by the team was bringing up 86 DOS
changing hardware environ-
o ltroit.ng. n"o"te the work had to be done in a
constantly
.nn'tg.s were also being made to the specifications of the budding operating
-"r.rt *frä"
system itself.

of the process, B6-DOS had to be compiled and integrated

s/ith the tslos' which
As pa the media Paterson's
frliJruroft *""i.tpin€i IBM to write, ancl this task was complicated-by (later
named
86-DOS not counting utilities such as EDLIN, CHKDSK' and INIT
pönllaf) - aiMicrosoft as one large assenbly-language program on an 8-inch
- "rriu.d
il"ppv airf. ff-t. rgM machine, however, uieci 5Yr-inch disks' so Microsoft
needed to de-
from
t..rr]ln. th. fot^"t ofthe new disk ancl then find a way to get the operating system
the old format to the new

Section I: The Deueloprnent ofMS DOS 75

 1980 tgal

Bi Gates (1982).

This work, handlecl by O'Rear, fell into a series ofsteps. First,

he movecl a secti()n ofcodc
from the s-inch disk and compiled it. Then, he converted the codeto
Inter hcxadecimal
format. Next, he uploaded it ro a DEC-2020 antl l.rom rheru clownlorded
it to a large Intel
fixed-disk developmenr system with an In,circuit Emulator. Thc DEc-2020,sed
for rhis
task was also usecl in clcveloping the BIOS. so there was aclclitional
work in <iownloading
the BIOS to the Intel machine, converting it to hexaclecimal format,
moving ir to an IBM
devclopment system, and thcn crossloading it to the IBM prototyp;.
Defining and imp)ementing the MS-DOS disk format different from paterson,s g inch
format-was an addcd challenge. paterson,s ultimate- goal for g6 DOS was
loSJical device
indepcndence, but during this first stage of developme-m, the operating
system simply ha.l
to be converted to handle logical records that were indepcnclent
of the physical record size.
Paterson, still with Seattle Complrter products, continued to work
on gb DOS and by the
cnd ofl9B0.had improved its logical device inclepenclence iry aclding
lunctions that
streamlined reading and wfiting multiple sectori and recor.l", *.ll o,
,".ords ofvariable
"" also worked
size. In addition to making sucl.r refinements of his own, paterson
on dozens
ofchanges requested by Microsoft, from modificatkrns to the ()perating
system,s startup
rressages to changes in EDLIN, the line eclitor he had written
for his own use. Throughout
this process, IBM's security restrictions meant that paterson was never
tolcl thc name ofthc
OEM and nevcr shown the prototype machines until he lefi Seattle Computer products
and
joined Microsoft in May 1981.

Änd ofcourse, throughout the process the developers encounte.ed the mydad
loose ends,
momentary puzzles, bugs, and unforeseen details withorlt which no project
is complete.
There were, for example, the serial card inteffupts that occurred whän
they shoulclnot
and, frustratingly, a hardwarc constraint that the BIos coulcl not
accon.rmoclate at first and
that resultcd in sporadic crashes during eariy MS DOS operations.

lil T h P I f S. t)us i n.),. t oped u

 1980-1981

Bob O'Rear's sketch of

the stery inuolued in
T*5*,.;*lr-.r fiouing 86-DOS to the
IBM Pratotyqe.
j_

!:
{

I
t
I
Ll
_r,/l.lL r^,'t"i
l"{l Ascvrcl*
f_
,(
t !1)
I /l I
--.-t-t
n v'Lk'd '1"
q M +v
-," *;.-at"-f/4J
la , tt, :-t+ "1i;
(
T. -'ffiirce
i-
u a0wr a-.'* ,)#^u
f;
r! ):';\r,.,_;.1y.'
t- " L")
9- ]ce ?g

J; c'"JL
e-"-il-"
I ( tte)k

I u-.Y^ lar tt-J.

Y
e

",'c^ ^P,.tr '{F,'a

(, sectian I: The Deuelopnent ofMS DOS 17

I
i
 1980-1981

t) ^ Dos clnr"as +t,res

a//- Pfie-/^ A,;* Qfl"-+,..' c* t"6, {-o!-al ,1.* 0^qre
q/*, t ,n [q F.dt to lo " e, *r La .jt-* * t^
/aAhn d;.t\ !*-x"r'6
^" + 4 U- y/{-''
+ fitfrnt -e otx: * b,e\ t- Lo . LJ *,: .tt
V' bt \2<ti,4ö\o. t6 DD\ ^*.) r. ro. . L" t"1ar
,/- 1s. clr^1, goe r. Fx-- t L-j tl* RI'Jr c Ms
Ä U-. .*..j ' s,-L-*l, *^t t-t+* t" t^ j; "
^
"
lJ. +\. *.N-. ^Ä"1,;4ä
; l^-;;

s c+--ys qG:""s
DD
&* t-a,; . so, -
t 9: ]

"'O@:tr-
qr:r, l'.& 1- BbDDs t" >-- aaL,e *J. o*l!- .
**^ f,-&
Flrl !.L 6t E6Dos r*-{ c. a.t r* ;;. ;--*tj
bhl'""dilt) to
- I

A ll"l'!-ri'q Lannttg ü'!A (o

ta naJ/. b" naro.rc. aa.r * , r
Aaro€r€..Ä+.r*J
ölfff *j \"* *{5J&E
e,,:+ 1i:l^.14. f_t",c+.q. -ää . ri+f "*
Orä"- vn J-t"*
rl€*J;"":-1ii.:"f *qi3#t
-H
"'. raw4qts *r"&l
'/-
, \df-";
./ ;+ d +o ezt,"'w:3' :E
"".a^ua
pfuffif o*-"-J" ar{^.Ie d"r"&{ b,41.,..1'

G üX+ f*
' ut--4;+l^
V ^".t e.r^l& z,eo 4uv sp^* -^. - "*
(O,.lu^ :q) r*ri. -- *uL-
^ ^ ^,.^0^
/. 1 cL.t "J es 2., o.+fJ ; - $ro\
rf' \f#cUr orJ J,rs'ür e,-----,
+ , J. c)*l n4 6r,"n ü^1 4 +L Loy ü.L. a-,:! t,
*n
0,e- Jup^& rn + ar u

a,/- k*Hl'donoc-;,.-r, a^" ,*,^r

r{. :r-I;.-z; .f.* c(rDr( o;r.UL A^"!"J aJÄ;"\
+

Palt ofBob O'Rear's "laundr!', Iist of opewting slstem changes and corrcctiofi.sfor
eartJ/ April 1981. Atoun.l
this time, interim beta copies uere shipped to IBM
f()r testing.

18 T he M S -DOS Enclc I opedia

 1981

The 1981 .lebut oI the

IBM Personal
'My ovm IBM comPuter
Imagine thatl'

Presenting the IBM of

Personal Computers.

'Dad, can I use

the IBM computer
tonight?'

In spite of such difficulties, howevel the new operating system ran on the prototype for
the iirst time in February 1981 In the six months that followed, the system was
continually
refined and expanded, and by the time of its debut in August 1981' MS-Dos' like the IBM
Personal Comiuter on which it appeared' had become a functional product for home
and office use.

Sectnn I t he Deuclapme/tt olMS'DOS \9

 1981

Version 1

The first release of MS-DOS, version 1.0, was not the operating system Microsoft envi
sioned as a final model for 16-bit computer systems. According to Bill cates, ,,Basically,
what we wantecl to do was one that was more like MS-Dos 2. with the hierarchicar fire
system and everything. . . the key thing lin developing version 1.0] was my saying, ,Look,
we can come ouf wirh a subset first and just go upward from that.,,'

This first version-cares's subset of MS-DOS-was actually a good compromise be_

tween the prescnt and the future in two important respects: It enabled Microsoft to meet
the development schedule for IBNI and it maintained progran translation compatibility
with CP/M
Available only for the IBM pcßonar compuret Ms-Dos 1.0 consistecl of 4000 rines of
assemblyJanguage source code ancl ran in g KB ofmemory. In acldition to utilities such
as DEBtlc, EDLIN, and FORMAT, it was organized into three major files. One
file,
IBMBIo.coM, interfaced with the RoM BIos lor the IBM pc and contained the disk and
character input/output system. A second file, IBMDOS.COM, contained the DOS kernel, in_
cluding the application program interlace and the clisk-file and memory managers. The
third file, COMMAND.COM, was the external command processor_the part olMS_I)OS
most visible to the user

To take advantage of the existing base ollanguages and such popular applications as
wordstar and dBASE Ir, MS-Dos was designed to allow software clevelopers to mechan-
ically translare source code for the 8080 to run on the 80g6. Ancl because oithis link,
MS-DOS looked and actecl like Cpr,M-8o, ar rhat time srill rhe standarcl among operating
systems for microcoülputers. Like its 8 bit relative, MS DoS useci eight character filenames
and threc-character extensions, and it had the same conventions for iclentifying clisk drives
in command prompts. For the most part, MS-DOS also usecl the same command language,
offered the same file services, and had the same general structure as cp/M. The resem-
blance was even more strikin6l at the programming level, with an almost one to_one cor
responclencc beiween Cp/M and MS-DOS in the system cails available to application
programs.

New Features
MS DOS was not, howevcl a CP/M lwin, nor.had Microsc.rft designed it to be inextricably
bonded to the IRM PC. Iloping to create a product that woulcl be successful over the long
term, Microsoft had taken steps to make MS,DOS flexible enough to accommodate
changcs and new-directions in thc hardq,'a.e tcchnology- disks, memory boards, even
microprocessors on which it depended. The first steps toward this inclependence from
-

20 T he MS-DOS Encyclopedia
 1981

ila -- l Big I.B.M.'s Little Computer

IBM's New Line Likely to Shake UP

The Market for Personal

ä=€+{tot
IBM Anno unce s Neu) M i ctocomput er E st e m
t s oJliciaL onc a,Pase

PFBSONAL COMPUTERS

PERSONAT
COMPUTER
FROM IBM
r,.itd.ntry iiro th. P.Eon.l
cohrulinl n.rk.l.in.lor
corror.t .r r.ll .t hom.

.4u(T*!'.Eah.n!q6o3lc4r

-'dFonü&'+ßvftld

A sampting ofthe heatllines and nellrspaper articles that abaun.led uhen IBM an\ounced its Personal

Section I: The Deüela?nent of MS DOS 21

t9f:u

A pageftom Microsoft s third-quarter

rePortfor 19€i1 .

IBM Breaks the

lSA Bärrier

specific hardware configurations appeared in MS-DOS vemion 1.0 in the form of device-
independent input and output, variable record lengths, relocatable program files, and a
replaceable command processor

MS-DOS made input and output device independent by treating peripheral devices as if
they were files. To do this, it assigned a reserved filename to each ofthe three devices it
recognized: CON for the console (keyboard and display), PRN for the printer, and ALX for
the auxiliary serial ports. rVhenever one ofthese reserved names appeared in the file con
trol block ofa file named in a command, all operations were directed to the device, rather
than to a disk file. (A file control block, or FCB, is a 37-byte housekeeping record located
in an äpplication's portion ofthe memory space. lt includes, among other things, the file
name, the extension, and information about the size and starting location ofthe file
on disk.)
Such device independence benefited both application developers and computer users.
On the development side, it meant that applications could use one set of read and write
calls, rather than a number of different calls for different devices, and it meant that an ap-
plication did not have to be modified if new devices were added to the system. From the

22 The MS DoS Enctclal,.edia

 1981

flexibility For example' even if a

user's point of view, device independence meant greater
programhadbeenclesignedfordiskl/oonly,theusercouldstilluseafileforinputor
direct outPut to the Printer.
Variable record lengths provided another step toward logical independence ln CPIM' logi-
cal and physical reärd ienS;ths were iclenrical: 129 bytes. Files could be accessed
only in
units ofi2g bytes and iile sizes were always maintained in multiples of 128 byres. with
MS DOS, howevcr, physical sector sizes were of no concern to the user' The ope'ating sys-
tem maintained filelengths to the exact size in bytes and could be relied on to suppolt logi-
cal records of anY size desired.

AnothernewfeatuleinMs-Doswastherelocatabieprogramlile'UnlikeCP/M,MS-DOS
haclthe,lbilitytoloadtwodifferenttypesofprogramfiles'identifiedbytheextensions
.CoM and .EiE. l'rogram files ending with .coM mimicked the binary liles in CP/M They
wcre more compact than EXE files ancl loaded somewhat faster, but the combined pro-
gram code, stack, and clata couid be no larger than 54 fB 't' nxl program, on the other
irand, could be much larger, because the cocle, stack, and data segments could be loaded
as

modules in separate parts of nemory determined by MS DOS Once the segments were in
set
memory, MS DOS then used part ofthe file header, or relocation rable, to automatically
the correct addresses for each of the different program segments'

In addition to supportin€i EXE iiles, MS-DOS made the external command processor'
just like any
COMMAND.COM, more adaptable by making it a separate reiocatable file
by custom command processoq as long
other program. It coulcl therefore be replacecl a
as the new file was also named COMMAND COM

Performance
Everyone familiarwith the IBM PC knows that MS-DOS eventualiy became the dominant
not
operating system on 8086-based microcomputers There were several teasons for this'
least of which was acceptance of MS-DOS as the operating system for IBM's phenomenally
successiul line of personal comp.tters. But even though MS-DOS was the
only operating
system available whcn the first IBM PCs wele shipped, positioning alone would not neces-
which appeared six months later'
sarily have guaranteed its ability to outstdp CP/M-86, .

US-öOS also oifered significani advantages to the user in a number of areas' including the
allocation and management of storage space on disk
however'
Like CP/M, MS-DOS shared out clisk space in allocation units Unlike CP/M'
the of these allocation units in a central file allocation table
MS DOS mappecl use
used a directory
-the
entry for
that was always in memory Both operating systems
recorcling inlbrmation about each file, but whereas a CP/M directory entry
included an
parts of the file
allocatio*n map-a list of sixteen 1 KB allocation units where successive
unit in the
were stored- an MS-DOS directory entry pointed only to the first allocation
FATandeachcntryinthetablethenpointecltothenextunitassociatedwiththefile'Thus,
a file
CP/Mmight require several dircctory entries (and more than one disk access) to load

sectian I: The Deuelopment afMS DoS 23

1981

larger than 16 KB, bur MS-Dos retainecl a comprete in-memory rist ofa file components
and all available disk space without having to access the crisk at all. As a result. l,t.s-oos,s
ability to find and load even very long files was extremely rapid compared wirh Cp/M,s.
Two other important features-the ability to read and writc multiple records with one
operating-system call and the transient use of memory by rhe MS_DOS command
processor provided further efficiency for both users and developers.
-
The independence ofthe rogicar record from the physical secror laid the foundarion for rhe
ability to read and write murtiple sectors. vhen reading multiple records in cp/M, an appri-
cation had to issue a read function ca[ for each sector, one ar a time. Vith MS_DOS, the
ap-
plication couid issue one read function call, giving the operating system the beginning
record and the number ofrecords to read, and MS DOS would then loacl all of rhe corrl-
sponding sectors automatical ly.

Another innovative featrrre of MS-Dos version 1.0 was the division of the command pro-
cessor, COMMAND.COM, into a resident portion and a transient portion. (There
is also a
third part, an initialization portion, which carries out the commands in an AUTOEXEC
barch file at startup. This part of coMMAND.coM is discarded from
memorv when irs
work is finished.) The reason for creating resident and transicnt porrions ol the co-mencl
processor had to do with maximizing the efficiency of MS-DOS fü the user:
On the one
hand, the programmers wanted COMMAND.COM to include commonly requested
func
tions, such as DIR and COpy, for speed and ease ofuse; on the other hanci, adding
these
commands meant increasing the size ofthe command processor, with a resuitingdecrcase
in the memory available to application programs. The solution to this trade off oT speed
versus utilitywas to inciude the extra functions in a transient portion of
COMMANb.COM
that could be overwdtten by any application requiring more memorv. To maintain the in_
tegrity ot rh(- [unction\ lor rhe u't-r. thc residenr prrt ;l COMM A\D.cOM * a. gi,
en the
job ofchecking the transient pofiion for damage when an apprication
termin"t.d. If n...,
sary, this resident portion would then load a new copy of its transient partner into
memory.

Ease of Use
In addition to its moves toward hardware indepenclence and efficiency, MS_DOS includecl
several services and utilities designed to make life easier for users and application devei-
opers. Among these services were improved error handling, automatic lo€lging of disks,
date and time stamping offiles, and batch processing.

MS-DOS and the IBM PC were targeted at a nontechnical group ofusers, ancl from the
beginning IBM had stressed the imporrrnLe ofdart inrcgrity Because data is most likely
to be lost when a user responds incorrectly to an error message, an effort was made to in_
clude concise yet unambiguous messages in MS DOS. To further recluce rhe risks of misin
terpretation, Microsoft used these messages consistently across all MS_DOS functions and
utilities and encouraged developers to use the same messages, where appropriate, in their
applications.

TheMS-DOS Encylopedia
 1981

Tuo pagesfrom Microsoft's MS-DOS üersion 1 O manual On the lefr, the slgem's requiremefits - 8 KB of
memory; on lhe right, the 118-page manual's complete table of contents

In a further attempt to safeguard data, MS-DOS also trapped hard errors such as critical
-
that had previously been left to the hardware-dependent logic Now
hardware errors
-
the hardware logic could simply report the nature of the error and the operating system
would handle the problem in a consistent and systematic way MS-DOS could also trap the
Control-C break sequence so that an application could either protect against accidental
termination by the user or provide a graceful exit when appropdate
To reduce errors and simplify use of the system, MS-DOS also automatically updated mem-
ory information about the disk when it was changed. In CPIM, users had to log new disks
asihey changed them-a cumbersome procedure on single-disk systems or when data
was stored on multiple disks. In MS-DOS, new disks were automatically logged as long as
no file was currently oPen.

Another new feature one visible with the DIR command-was date and time stamping
-
of disk files. Even in its eafliest forms, MS-DOS tracked the system date and displayed it at
every startup, and now, when it turned out that only the first 16 bytes of a directory entry

Section I: The Deuelopment ofMS DOS 25

-_:ti 1982

were needed for file-header information, the MS_DOS programmers

decided to use some
ofthe remaining 16 bytes to record the date and time ofcrieation or update (and
the size of
the file) as well.

Batch processing was originally added to MS-DOS to help IBM. IBM wanred
ro run
scripts sequences of commands or other operations one after the
- ofthe system. To do this, the
functions
other to test various
testers needed an automated method ofcalling
routines sequentially. The result was the batch processor, which later
also provicled uiers
with the convenience of saving and running MS-DOS commands as batch
files.
Finally, MS-DOS increased the options available to a program when it terminared.
For ex
ample, in less sophisticated operating systems, applications and other programs
remained
in memory only as long as they were active; when terminated, they were
rem<>ved from
memory MS-Dos, however added a terminate-and-stay-resident function that
enabled a
program to be locked into memory and, in effect, become part
ofthe operating_system
environment until the computer system itself was shut clown or restarted.

The Marketplace
When IBM announced the personal Computer, it said that the new machine
would run
three operating systems: MS-DOS, Cp/M-g6, and SofTech Microsysrem,s p_Sysrem.
Of the
three, only MS-DOS was available when the ItsM pC shipped. Nevertheleis,
when MS_DOS
was released, nine out often program s on the InfoWoriA bestseller
list for 19g1 ran under
CP/M-80, and CP/M 86, which became available about six months later,
was rhe operating
svstem ofchoice to most writers and reviewers in the trade press.

Lnderstandably, MS-DOS was compared with CplM 80 and, lateq Cp/M_g6.

The main con_
cern was compatibility: To what extent was Microsoft's new operating
system compatible
$'ith the existing standard? No one could have foreseen that MS DOS
would not oniy catch
up with but supersede CP/M. Even Bill Gates now recalls that ,,our
most optlmistic vies,. of
the number of machines using MS DOS wouldn,t have matched what
realiy ended up
:uppening."
To begin with, the success ofthe IBM pC itself surprised many industry
watchers. \:Vithin a
1'ear IBM was selling 30,000 pCs per month, thanks in large part
to a business community
:hat was already comfortable with IBM's .,ame and reputation and,
at least in retrospect,
.r.as ready for the leap to personal computing.
MS DOS, ofcourse, benefitecl enormously
::om the success ofthe IBM pC-in large part because IBM supplied all its la.rg.,ug.,
,lpli.ations in Mq-DOS formar "rd
3-rI. at first, wrirers in rhe trade press still believed in cp/M and questioned the viabiiity
of
-:. rel1' operating system in a world dominated by Cp/M g0.
Many assumed, incorrectly, that
-: CP \1-86 machine could run CplM-8O applicarions. Even before Cp/M g6 *as
a,railaüie,
a1:ure Computing referred to the IBM pC as the ,,Cp/M Record plaver,,_ oresumabiv in
,:i::cipation of a vast inventory of Cp/M applications for rhe new computer_ and
lej its
::-:iers to assume that the pC was actuailv a Cp/M machine

-..',: :.Da\ Entyloped n

 1981 1982

or any other
Microsoft, meanwhile, held to the beliefthat the success of IBM's machine
the emergence of an industry standard for a
16-bit microcomputer depended ultimately on
could not afford to develop software for even
16-bit operating system. Software developers
coulcl (or would) not pay the the
two or three difterent operating systems' and users Prices
would almost certainly
developers would have to charge if they did Furthermore,-users
rebel against the inconvenience of sharing data stored underdifferent operating-system
to be
formati There had to be one operating system, and Microsoft wanted MS-DOS
the one.
hardware
The company had aheady taken the first step towarcl a standard by choosing -
portability' and
indepenäeni designs wherever possible Maihine inclependence meant
port;bility meant ihat Microsoft could sell one version of MS-DOS to different hardware
alone'
manufacturers who, in turn, couid aclapt it to their own equipment Pofiability
the standard'
however, was no guarantee of industry wide acceptance' To make MS-DOS
for MS-DOS And in
Microsoft needed to convince software developers to write programs
1981, these developers were a little confused about IBM's new
operating system

An operating system by any other name...

A tangle of names gave rise to one point ofconfusion about MS DOS' Tim
Paterson's
"Qui& ancl Dirty Operating System for rhe 808b was originaliy shipped by Seattle
s6-ood efter trlicrosoft purchased 8b-DoS' the name remained
cämpnt", rtodu.ts as
the time the was ready for reiease, the new system was known as
for a while, but by PC
MS-DOS' Then, after the IBM PC reache<l the market, IBM began
to refer to the operating
press soon shortened to
system as the IBM Personal Computer DoS, which the trade
and DISKCOMP. that
PC_DOS. IBM,s version conrainea some urilities, such as DISKCOPY
by other manufac-
were not inclucied in MS-I)OS, the generic version available for license
to the confusion about
turers. By calling attentlon to thes;ifferences, publications added
the distinction between the Microsoft and IBM releases of MS-DOS'
promote MS-DoS but
Fu her complications arose when lifeboat Associates agreed to help
86 rhus became one of a line
decidecl to call the operating system bftware Bus MS-DOS
oitrademarkecl Software Bus products, another of which was a product called SB-80'

Lifeboat's version of CP,/M-80.

wanted to use their
Finaliy, some of the first hardware companies to license MS-DOS also
names as
own names ior the operating system Out of this situation came such additional
COMPAQ DOS and Zenith's Z-Dos
could become the industry
Given this confusing host of names for a product it believed
insisted that the operating sys-
stanclard, Microsoft iinally took the leaci and, as developel
tem was to be called MS DOS Eventually, everyone but IBM complied

Developers attd MS-DOS

of Microsoft's business-
Early in its career, MS-DOS represented just a smallftaction
languages ln addition' in the
much la4aer revenues were generated by BASIC and other
growth of CP/M-86 and other
first two vears after the intr;duction of the IBM PC, the

Sectianl: The De elc)pment of MS'DOS 27

 1981 1982

environments nearly parairelecl that of MS-Dos.

&r Microsoft found itselfin the unenviable
positbn of giving its suppofi to MS Dos-while arso selrrng languages to ,.,n on cplM_g6,
thereby contributing to the growth ofsoft*ar.
fo. ius_oösls f,rgg!",.o_p",r,o.
Given the uncertain outcome ofthis two horsc race,
some other software developers
j:_:^ll which way the ha.awa,e manufaliu..r, *"Jro
:l::: la-:ee
r ne nardware manutacrurers
i.,_p u"r their pan,
were confronting the issue of compatibiiity tetween
ojerat-
l:* *aT1 Spe-cifically, they needed to be cJnvin..a,n"t uS_föi *as nor a maverick_
that it could perfcrrm as well as CplM_g6 as a_ base
for appti."tlo^if.,nt f.,oa freen porred
from the CyM-80 environmenr for use on t6_bit compuLrs.
Microsoft approached the problem by emphasizing four
relateci points in its cliscussions
with hardware manufacturers:
o Fi.st, one ofMicrosoft,s goals in cleveloping the first
version of MS_DOS had always
been rranslation compatibility from CpiU_äO to US_OOS
-
' second, tralslation was possible only for software
wrrtten"oit*",.
in g0go or zg0 assembly
language; thus, neither MS DoS nor Cp/ru 86
could run profai]s written tbr other
8-bit processors, such as the 6800 or rhe 6502.
o Third, many applications were written in a higltJevel
langgage, rather than in assem_
bly language.
a Fouth, most ofthose highJevel langua€ies were
Microsoft products and ran on
MS DOS

Thus, even though some people had originally believed

rhat only Cp/M_g6 would auto
matically make the installecr base of cplM g0 software
available i.ile tg^4 pc ancr other
16-bit computers, Micr<tsoft convinced the
harclware manuf".,rl..r, ih", us DOS was, in
actuality,as flexible as Cp/M-s6 in irs compatibility wirh
existingl ana appropriate_
CP/M-80 software.

MSJ)os was put at a disadvantage in one area, however, when

Digital Research convinced
several manufacturers ro include both 8080
ancl8086 chips in i'ei?ilactrines. lrirh s_bit
and 16-bit software used on the same machine, the
user cou[ ..ry or, ,fra ,"a. disk format
for both rypes of software. Because MS_DOS used
a different disk format. CplM had the
ed-fae in these dual-processor machines _
although, in fact, it did noiseem to have much
effect on the survival of Cp/M-86 after the first year
or so.
Although making MS-DOS the operating system of obvious preference
was not as easy as
simply convincing hardware manufacturers to off'er
it, Microsorrstist or us_oos custom_
ers grew steadily from the time the operating
system was introduced. Many manufacturers
."":l"u:9 gff.t along tr ith MS DOS, but by rhe cnd of1983 rhe rechnical
_,: 9p/M-86 supe_
riority of MS-Dos (bolstered bv rh e introcluction of such products
as Lotus 1-2 3) carried
the market. For exampre, when DEC, a rongtime horclout,
iecicled to make MS Dos rhe pri
mary operatinla system lor its Rainbow computer, the company
mentioned the dcher set of
comnrands and "dramatically" better disk performance
of'MSIDOS as ,easons f,rr" it.
( hr)rce (,\ cr CP M 8()

28 The MS DOS Encycbf,edia

 19811982

Mfi€M@6@FF

original equip/nent manufacturer (oEM) marketing hrochure describing the slrengths of MS DOS
A Microsofi

Section I: The DeueloPmeü 1fMS'DOS 29

 1982 1983

Version 2

After the release of pc-specific version 1.0 of MS-DOS, Microsoft worked on an update
that contained some bug fixes. Version 1.1 was provided to IBM to run
on the upgraded pC
released in 1982 and enabled MS DOS to work with double_sided,
320 KB flopiy ai*".
This version, referred to as 1.25 by all bur IBM, was rhe first version
of MS_Död inlppeO ly
other OEMs, including COMpAe and Zenith.

Even before these intermediate releases were available, however,

Microsoft began plan_
ning for future versions of MS-DOS. In developing the first version, the programmers
had
had two primary goalsr running translated Cp/M g0 sofr*are nnci
keeping MS_DOS small.
They had neither the time nor the room to inclucle more sophisticated
features, such as
those typical of Microsoft,s UND< basecl multiuser, multitasking
operating system, XENIX.
But when IBM informed Microsoft rhar rhe nexr major edition
of tie pC would be the
Personal Computer XT with a 10-megabyte fixed disk, a larger,
more powerful version of
MS DOS-one closer to thc operating system Microsoft had
envisioned fiom the stafi_
became feasible

There were three particular areas that interested Microsofi: a new, hierarchical
file system,
insullable device drivcrs, and some type of multitasking. Each of these leatures
contrib_
utcd to version 2.0, and together they represented a major change in
MS DOS while still
n1aintaining compatibility with version 1.0.

The File System

Pri.rary responsibiiity for version 2.0 fell ro paul Allen, Mark Zbikowski, and Aaron
Reynolds, who wrote (and rewrote) most of the version 2.0 cocle.
The major design issue
confronting the developers, as well as the most visible cxample of its difference
from ver_
sions 1.0, 1.1, and 1.25, was rhe introduction of a hierarchical hle
sysrem ro handle the fiie_
management needs ofthe XT,s fixed disk.

Version 1.0 had a single directory for all the files on a floppy clisk. That system
workecl well
enough on a disk of limited capaciry, bur on a 10-megabyie fixed clisk a single
directory
could easily become unmanagcably large ancl cumbersome.

CP/M had approached the problem of high-capacity storage media by using partirioning
a
scheme that divided rhe lixed disk into 10 user areas equivalcnt to 10 separarelloppy
disi
drives. On the other hand, LINIX, which had traditionally ciealt with larger systems,
used
a branching, hierarchical file structure in which rhe u"er eould creare
directories ancl
subdirectories to organize liles and nake them readily accessible. This was the file_
mana€aement system implemented in XEND(, and it was the MS_DOS
team,s choice for
handling files on the XT's fixed disk.

30 7 hc 145 DOt EtlL)tclape.ha

 1982 1983

The MS DOS uersion 1.O manual next to the rersion 2.0 nanüal

Partitioning, IBM's initial choice, hacl the acivantages o1 familiarity, size, and ease of implc
mentation. Many smali system users- particularly software developers-wcre already
familiar with partitioning, if not overly loncl of it, from their cxperience with CP/I4' Devel
opment time was also a maior concern, and the code necded to devel 'p J fartitioning
scheme would be minimal comparecl n'ith the cocle required to manage a hierarchical
filc
jmc to implcment'
system. Such a scheme would also take less t

However, partitioning had two inherent disadvantafles First, its functionality would
decrease as storage capacity increased, and even in 1982, Microsoft $'as anticipaling sub-
stantial growth in the stQrage capacity of disk based media. Second, Partitjoning de-
pendeci on the physical clevice. If the size oi the disk changed, either the number or the
iize ofthe partiiions musr als. be changed in the code tbr both the operating systcm and
the applicaiion programs. For Microsoft, with its commitment to hardware independence'
partitioninll would have tepresentecl a step in the \\"rong direction
A hierarchical file strr,rcture, Qn the other hand. could be independent i)f the Physicel
device. A clisk could be partitionccl Jogically' rather than physically And because these
partitions (directories) were controlled by the uscr, they wcre opcn ended and enabled
the individual to determine the best way of organizing a disk

Ultimately, it was a hierarchical file system that found its way into MS l)OS 2 0 and even-
tually convincecl every()ne that it was, indccd, the better and more flexible solution lo the
problem of supporting a lixed disk The 1ile systen was logically consistent with the
IENX fil. yet physically consistent with the file access incclrporated in versions
"tt*rure,
1.x, and was based on a root, or main, directory under which the user ccmld create
a sys

tem of subdirectories and sub subdircctorics to hold files. Each filc in the system was iden-
tifiecl by the directory Palh leaclinll to it, and the number of subdirectories was limjted only
by the length olthe pathname, which coulcl not excecd 64 characters'

In this file structure, all the subdirectories and the filename in a path wcre sepanted
from one another by backslash characters. which represcnted the only anomaly in the
xENIX/MS-DOS systcm of irierarchical files. XENIX used a fbrward slash as a sepalator.
systems'
but versi()ns 1.x of MS-DOS, borro*'ing fiom thc tradition of DEC oPerating
at IBM's
already usecl the forrvard siash for switches in the commancl iine, so Microsofl'
request, deciclecl to use the backslash as the seParator instead Although the backslash

Section I: The DeoeLapm.ent o.l'x'15 DOS 31

 10cl 198J

character created no practical problems, except on keyboards

that lacked a backslash, this
decision did introduce inconsistency bet*...r us-oo.s existing uNIXJike op.."ii.,g
by Äabling th. .,r., t,., "
"nd
systems. And although Microsoft solved the keyboard pr'blem
change the switch character from a srash to a hyphen, ihe solution
itselfcreatecr compati,
bility problems for people who wished to exchange batch files.
Another major change in the file_management system was reiated to
thc new directorv
structure: In Order to fully exploit a hierarchical file System, Microsoft had
to add a new
nay of t allr ng file.ervices
versions 1.x of MS-Dos used cp/Mlike structurcs callecr firc control
blocks. or FcBs. ro
maintain compatib ity with order cp/M g0 programs. The FcBs !onrained
alr perrinent
information about the size and loc:rrion of a file bur did not alow the user
tQ specify a file
in a different directory. Therefore, version 2.0 of MS Dos needed
the ada"d atility to ac
cess files by means of handles, or descriptors, that could operate
across directory lines.
In this added step toward logical device independence, MS Dos
.eturned a hancile when
ever an MS-DOS program opened a file. All further interaction
with the file involvecl only
this handle. MS-DOS lrade all necessary acljustments to an internal
structure _ diflerent
from an FCB-so that the program never had to cleal clirectly with
informarion about the
file's location in memory. Furthermore, even if future versions
of MS_DOS were to change
the structure ofthe internal control units, program code would
not need to be re*.itteri_
the file handle would he rhe only referent ,,eeJ".l, and rhis would
not change.
Putting the intemal control units under the supervision of MS DOS
and substituting
handles for FCBs also marle it possible for MS_DOS to redirect a program,s
input ancl out
put. A system function was provided that enabled MS DOS to
divert the reads or writes
directed to one handle to the file or clevice assigned to another handle.
This capability was
used by coMMAND.coM to allow ourpur from a fire ro be reclirected
to a de't., ,.,.h
p nter, or to be piped to another program. It also allowed system cleanup on progrant ". "
term inat ir ,ns.

Installable Device Drivers

At the time Microsoft began developing version 2.0 of MS_DOS, the
company also realizccl
that many third-party pcripheral devices werc not working well with
one another. Each
manufacturer had its own way of hooking its hardware in; MS_DOS anci if
two third pafiy
devices were piugged into a computer at the same time, they woulcl
often conflict or flil.
One ofthe hallmarks of ITiM's approach to the pC was open architecture,
meanin€a that
users could simply slide new cards into the computer whenever
new input/outpuicle_
vices. such as fixed djsks or printers, were added to the system.
Unfortunately, iersion
1.0 of MS-DOS did not have a corresponding open architecture
built inro it_rhe BIOS

l2 T he,U S DOS EncJ)clopedia

 Da2-1983

contained all the code that permitted the opcrating system to run the hardware lf inde
pendent hardware manufacturers wanted to develop equipment for use with a computer
manufacturer's operating system, they would have to either completely rewrite the device
drivem or write a complicated utility to read thc existing drivers, alter them, add the code
to suppo.t the new device, and producc a working set of drivers. Ifthe user installed more
than one device, these patches would often conflict with one anothcr' Fufihermore, they
would have to be revised each time the computcr mlnufacturer updared its version
of MS-DOS.

By the time work began on version 2.0, the MS-DOS team krew that the ability to instali
any device driver at run time was vital. They implemented installable device drivers by
making the drivers more modular. Like the FAI, IO.SYS (IBMBIO.COM in PC DoS)
became, in effect, a linked list this time, of device drivers could be expanded
-that
through commands in the CONFIG.SYS file on the system boot clisk. Manufacturers could
now write a device ddver that the user could install at run time by including it in the
CONFIG.SYS file. MS DOS could then add the device driver to the linked list.

By extension, this ability to install device drivers also added the ability to supersede a pre
viously installed driver-fbr example, the ANSLSYS console ddver that suppo.ts the ANSI
standard escape codes for cursor positioning and screen contr()I.

Print Spooling
At IBM's request, version 2.0 of MS-DOS also possessed the unclocumented ability to per-
form rudimentary backgror-rnd processing an interim solution to a growing awareness of
the potentials ef multitasking.

Background print spooling was sufficient to meet the needs of most people in most situa-
tions, so the pdnt spooler, PRINT.COM, was designed to run whenever MS DOS had
nothing else to do. \Vhen the parent application became active, PRINT COM would be in-
terrupted until the next lull. This type ofbackground processing, though both limited and
extremely complex, was exploitecl by a number oiapplications, such as SideKick

Loose Ends and a New MS-DOS

Hierarchical files, installable device drivers, and print sprxrling were the major design
clecjsions in version 2.0. But there were dozens of smaller changes, too

For example, with the fixeci disk it was necessary t() modify the code lor automatic logging
ofclisks. This modification meant that MS-DOS had to access the disk more often, and file
access becamc much slower as a result. In trying tofind a solution to this problem, Chris
Peters reasoned that. if MS-DOS had iust checked the clisk, there was some minimum time

Section I: The l)eoelapme ,|MS-DOS 33

1982-1983

Tüomembers aJ the
IBM line ofpersonat
camputctsfor uhlch
Dersi.r sland2of
MS-DOS uere cletel
oped. on the leJi, the
oriBinat IBM PC (rer-
sio 1.O of MS-DOS\
on the ri8ht, the IBM
PC|XT(Lersian 2.O).

lSAgt

a userwould necd to physically change disks. Ifthat mininum tine had not elapsecl, the
curfent disk information in RAM whether for a fixed disk or a flQppy-was probably
still geod.
Peters lound that thc fastest anyone could physically ctrange clisks, even if the disks were
damaged in the process, was about tw. seconds. Reasoning from tl.ris observation, he ha<1
MS-DOS check to sce hoq' nruch time had gone by sincc the last disk access. If less than
two seconds l.rad elapsed, he had MS DOS assume that a new disk hacl not been inserted
and that the clisk informarion in RAM was srill valid. \trith tbis littlc rrick, rhe speed of file
handling in MS-DOS version 2.0 increased considerably.
Version 2.0 was released in March l98l. the product of a surprisingly smail team of six cle_
velopers, including Peters, Mani Ulloa, and Nancy panners in addition to Allen, Zbikowski,
and Reynolds. Despite its conplex new fcarures, version 2.0 l,as only 24 KB of code.
Though it maintained its conpatibiiity with vcrsions 1.x, ir was in rcaliry a vasrly differenr
operating system. \J(/ithin six months of its release, version 2.0 gained I,idcspread public
acceptance. In acldition, popular application programs such as Lotus 1 2-3 took advanta€le
of the f'eatures ofthis new version ofMS DOS and thus helped sccure its luture as the
industry standard for [1086 processors.
Versions 2.1 and 2.25
The worlcl into which version 2.0 of MS-DOS emcrged w?s consiclerably different from the
one in which version 1.0 made its debut. !(hen IBM released irs original pC, the business
market for microcomputers was as yet uncielined- if not in scope, at least in terms ofwho
and what would clominate the field. A year and a half later, when the pC/XT came on the
scene, the market s,'as much better kr]own. It had. in fact, been heavily infh,renced by IBM
itself. There were still n-rany MS DOS rrachines, such as the Tandy 2000 and the Hewlett
Packad HP150, that were hardware incolnpatible with the IBM, but manufacturers of new
computers knew that IBM was a force to consider and many chose to compete with the
IBM PC by emulating it. Software clevelopers, too, had gaincd an understanding ofbusi
ness con]putinEl and were conlident thcy could position their soltware accurately in the
enormous MS-DOS market.

34 The MS DAS Enclclapedia

 1983

In such an environment, concerns about the existing base of CP/M software faded as
developers focused their attention on the fast-growing business market and MS-DOS
quickly secured its position as an industry standard. Now, with thc obstacles to MS DOS
diminished, Microsoft found itself with a new concern: maintaining the standarcl it had
created. Henceforth, MS-DOS had to be many things to many people. IBM had require
ments; other OEMS had requirements. And sometimes these requirements conflicted

Hardware Developers
When version 2.0 was released, IBM was already planning to introduce its PCjr. The PCir
would have the ability to run programs fuom ROM cartridges and, in addition to using half-
height 5%-inch drives, would employ a slightly dilferent clisk-controller architecture. Be-
cause ofthese differcnces from the standard Pc line, IBM'S immediate concern was lor a
version 2.1 olMS-DOS modified lor the new machine.

For the ionger term, IIIM was also planning a faster, more powerful PC with a 20-megabyte
fixcd disk. This prospect meant Microsoft nceded to look again at its file-management sys-
tem, because the larger storaEle capacity of the 20-megabyte disk stretched the sizc limita-
tions for the file allocation table as it worked in version 2 0

However, IBM's primary interest for the next major release of MS DOS was networking
Microsoft would have preferred to pursue multilasking as the next stage in the develop-
ment of MS-DOS, but IBM was aheady developing its IBM PC Network Adapter, a plug-in
card with an 80188 chip to handle communications so as soQn as version 2 0 was released,
the MS DOS team, again headed by Zbikowski and Reynolds, began work on a networking
version (3.0) of thc operating system.

Meanwhile...
The international market for MS-DOS was not significant in the first lew )€ars after thc
release of the IBM PC and version 1.0 of MS DOS IBM did not, at first, shiP its Personal
Computer to Europe, so Microsoft was on its os'n there in promoting MS DOS In 1982, thc
company gained a significant advantage over CP/M 85 in Europe by concluding an agree-
ment with Victor, a software company that was very successful in Europe and had already
licensecl cP/M-86. vorking closely with victor, Microsolt provided speciai development
support for its graphics adaptors and eventually convinced thc company to offcr its pro-
ducis only on MS DOS. InJapan, the most popular computers were 280 machines, and
given the country,s huEle installed base of 8$it machines, 16 bit computers wele not taking
hold. Mitsubishi, however, offered a 16-bit computer. CP/M-86 was Mitsubishi's
^lthough
original choice for an operating system, Microsoft helped get Multiplan and FORTRAN
running on the CPIM-86 system, and eventually won the manufacturer's support for
MS-DOS,

Section I: '1he DeteloPmetLt ofMS DOS 35

 1983

A sample o.f the reDieus that appeared

ui lh e.1ch fieut uersion of MS-DOS .

Inesistible
DOS 3.0
hln).t 4J vr.n tklv,{4t
ro ßt aa ri,üa;4 ,,t-"..,1

The fucent
Of DOS

MS-DOS 2.00: A
Hands-OnTütorial

In the software arena, by the time cle\€lopment was unclerway on the 2.x
releases of
MS-DOS, Microsoft's other customers were becoming more vocal about
their own needs.
Sevcral wanted a networking capability, adding weight to IBM,S request, but
a more urgent
need for many- a need nol shared by IBM at the time support for intcrnational
-wasof us Dos that
products. specifically, these manufacturers needed a versio''
could be soki
in other corintries a version of MS-DOS that could display messages in other
languages
and adapt to country-specific conventions, such as date and time fcxmats.

Microsoft, too. wanted to internarionalize MS-DOS, so the MS DOS team, while

mo<lifying
the operating system to support the pcjr, also added functions ancl a couNTRy
command
that ailowed users to set the date and time formats and othcr country
dependent variables
in the CONFIG.SYS file.

36 The,14 S DOS Enc).topcdß

 1983

A Kanji screen üith

the MS DOS coPtright
NEC Pc 9300 Serie.

?1rd)rf MS-DOS 3 10
^',-r'r'
copyrieht r931,1935 Microsort Corp / NEc corpofiio^

iSani4j nri*.BqfE<f
;+*tl. , r 7 l. H:t'1lo NEcorc sYS <!
cc[ilaNo ,'!, 3. ]0
^'
f -1a at4^raiit)=-Ät{rL'li IAIAI RYL
^,
7iv?r)lt^:.{BtN
'F^D.^ c^r .a,-/ c{ a\<l l\ cclJ 0l B
'- lor(0o"
Yq
0? I :rt 'cv
i.nr :oi tr'ro :(, üt] ts -o/
n ßa7 l4btt Dtt
r6oaaso I L' l4rql"lhe" f,
)drt rlrr$"+}i
^,?t
R tn'rjl iq+M s Do s

At about the same time. another intefnatiooal requirement appeared. TheJapanese

malket
(icleo-
for MS-DOS was growing, and the qucstbn ol supporting 7000 Kanji characters
bytc characters For English
grams) arose. Th; diffi;lty with Kanji is that it requires dual
byte cQrresPonds to one characLer Japanese char-
änd most European character scts, Qne
sometimes t\!'o This variability creates prob
acters, however, sometimes use one byte,
a string f|om the
lems in parsing. ancl as a result MS-l)oS had k) be modiiied to parse
beginning, rather than back up one character at a time
ol MS-DOS
This support lor individual country formats and Kanji appearecl in version 2 01
IBMclidnotwaotthisversion,soijupportfbrthePCjr',developeclbyZbikowski'Reynolds'
Ulloa, and Eric Evan.' npp.n..d ,.p'iately in version 2 l which *'ent only
ttl IBN'I and did
not include the modifications lor international MS-DOS

Different customers, different versions

As early as version 1.2i, Microsoft face<1 the problem of Lrying to satisfy
those OEM ct'ts
tome6 that wante.l to have the same version of MS-DOS as IBM Some such as COMPAQ'
any differ-
were in the business of selling 100-pe|cent compatibility t!'ith lBNl' For them'
possibility of
ence between their ve.sion Jth. operating system ancl IBM
s introduced the
was noi until ver-
incompatibility. Satisfying these requests \"as dilficult' ho\\'ever' ancl it
Microsoft was able to supply a systel.n that other OENIs agreecl was identicai
sion 3.j that
with IBM'S.
vcrsions 2 1 and 2'01 tcr
Before then, to satisfy thc OEM cr'rstonrers, Microsofi con.rbincd
this because ()frhe internirtionalizetion
create venion 2.11. Although IBM did n()l accept
version fol all non-IBM !ustolncrs rLrn n ing an)
cocle, vcrsion 2.11 became thc standard
$as sold n'c.rrld*ide and rranslated into
form of MS DOS in the 2 x serics. Vcrsion 2 11
internlediate versions provi'led support lor
about 10 different languages Two other
Hangeul (the Korean charactcr set) ancl Chinese Kanji'

Sectic)n I The DeLc[aPme t o.f]IS DOS 37

 Software Concerns
After the release ofversion 2.0, Microsoft aiso E{ained an appreciatien ofthe importance_
and cliiliculty-of supporrin{a rhe people who were devel-oping software for
M.S_DOS.
Soft*.are developers rvorried about clownward compatibiiity.
They also worriecl about
upward con.rpatibility. But despite these concerns, they sometimes
used programming
practices that could guarantee neither. When this happenccl
and th. res.,ltin! prograÄs
wcre successtirl, it was up re Microsoft to ensurc compatibility.
For example, bccausc the infofmation about the intcrnars
ofthe BIos and the RoM inter-
face had been published, sottware clevelopers coulcl, ancl
often clid. work clirectly with the
hardwarc in order to get nore speed. This meant sidestepping the
operating system for
some operations. However, by choosing to work at the lowc.levels,
ihese de,relopers lost
tlte protection provided by rhc ,rpe rarj;g 5!srcnt asainst hardware
cltanges. Thus. when
l<n. lcvel clnnges were madc in the hardware, theü program,
.irh., .il;;,;.;k ,;;;
.ol rLn c( 'op(rdti! el1 n ith , ,rhr r appiic.rr iun.
Another softrr,arc problem was the continuing neccl for compatibility
with Cp/M. For
example, in cP/M, programncrs w.uld ca a fixecl acrclress
i" r"* ..-.ry in oJ., ,o ..
quest a functionr in MS DOS, they would requcst operating
systcm services by executing a
sofiware interrupt. To support older softwarc, thc first version of
MS_DOS all r*.d o p.,i
gram to request functions by either otcthod. One of the
Cp/M_based program, ,,,ppolr.d
in this fäshion was thc very popular \üorclsta. Since Microsoft
could not make changes in
MS,DOS th.rt l,I)uld makc ir impossible to run such a I,iclely
usecl program, each nei ver_
sion of MS-DOS had to continue supporting Cp/M_style calls.

A more pervasive ClTM-related issue n-as tlie usc of FCB

style calls for file and record
managenrenr. The version 1.x rcleases ofMS DOS had used FcB_style
calls exclusively, as
had cP/M Version 2 0 inrroduced rhc more efficient and flexibre
handre cats. bur Microsofr
could not simply abolish the old FcB_style calls, because so many popular
programs used
rhcn rn fäct. s.n-re of Micro5efl'5 qa'p languages used the-. s", ir's öos r.'"Jä,.,pp.r, "
both types of calrs in the vcrsion2.x serie". To .ncourrg. rhe use of
the new hancne^cars,
however, Microsoft macle it easy for MS DOS users to r,rigrade
to vcrsion 2.0. In acldition,
rhe company convinced IBNI ro rcquire versidn 2.0 for Lh! pC,/XT
ancl also encouraged
softwarc developers to require 2.0 lor their applications.
At first, both softwarc deveiopers ancl OEM customers were reluctant
to require 2.0
because they were concernecl about problems with the installed
user base of 1.0
systems requiring version 2.0 neant supporting both sets ef calls. Applications
-
neeclecl to be able to detecr wlri. h version , rf rhe oper.rting .ysrcm
also
the uier was running.
For versi()ns 1.x, the programs nould have to use FCB call.s; for versions
2.x, they would-
use rhe file har]dles ro exploit rhe flexibility of MS_DOS more
lully.
All rold. it was ar.) awkward periocl o| transition, but by the time Microsoft
began work on
lcrsion 3.0 .rnd the support for TBM'. upL(,ming 20 m;grbyre fixed
disk, it hacl become
apparent that the change hacl been in everyone,s best inte;esr

38 '1
he rVS DOS Encyctope,tia
 1983-1981

Version 3

The types of issues that began to emerge as Microsoft worked toward version 3 0, MS-DOS
for networks, exaggerated the problems of compatibility that had been encountered
before.

First, networking, with or without a multitasking capability, requires a level ofcooperation

and compatibility among programs that had never been an issue in earlier versions of
MS-DOS. As described by Mark Zbikowski, one of the principals involved in the proiect,
"there was a very long pcliod of time between 2.1 and 3.0 almost a year and a half. Dur-
ing that time, we believed q'e understood all the problems involved in making DOS a net-
working product. IButJ as time progrcssed, we realized that we didnl fully understand it,
either from a compatibility standpoint or from an operating-system standpoint We knew
very well how it [DOS] ran in a single tasking environment, but we started faoing to this
new environment and found places where it came up short."
In fact, the great variability in programs and programming approaches that MS-DOS
supported eventually proved to be one ofthe biEigest obstacles to the development of a
sophisticateci networking system ancf, in the longer term, to the 2.ldition oftrue
multitasking.
Fufiher, by the time Microsofi began work on version 3.0, the programming style of the
MS-DOS tearr had changeci considerably. The team was still small, with a core group of
just five people: Zbikowski, Reynolds, Peters, Evans, and Mark Bebic. But the concerns for
maintainability that had dolinated Programming in larger systems had percolated down
to the MS-DOS environment. No$ , the desire to use tricks io optimize for speed had to be
tempered by the need for clarity and maintainability, and the small package of tightly
written code that was the early MS DOS had to be sacrificed for the same reasons

Version 3.0
All tolcl, the work on version 3.0 of MS-DOS proved to be long and clifficult. For ayear and
a half, Microsofi grappled with problcms of software incompatibility' remote fiie manage-
ment, and logical device independence at the network level Even so, whcn IBM was ready
to announce its new Pcrsonal Computer AT, the nelwork software for MS-DOS was not
quite ready, so in August 198'1, Microsoft released version 3.0 to TIIM without network
softwarc.
Versi()n 3.0 supported the AT's larger fixed disk, its new CMOS clock, and its high-capacity
1.2 megabyte lloppy disks. It also providcd the same internatiQnal support included eadier
in versions 2.01 and 2.11.'fhese featurcs q,'ere made availablc to Microsoft's other oEM
customcß as version 3.05.

Sectian I: The Deuetopment DOS 39

'JMS
1983-1984

f";ö
Yv

Roq

9L

Aaron Rqnaus's diagram of rcrsbn 3.O's nehoork sltpport, sketched out to enable
him k) add the-fail option
to Interrupt 24 and./ind al/ ptaces uhere existing parts
of MS-DOS uere affectetl. Eren after netu;rking haLl
become a realiry, Retnotds kept this diagram pinned to his alfice
ua// siiply because ,.rt uas sa much u.)rk

The trs DOS Enclclopedia

 1983-198l

Röt^n
u'r^,- n r. 14-
F:r.r.
r .45 ri:alI
Jl-_/
i4 k1bö€ )

] o,o.r,.
c,rosE

I *un* t,.
j- I

( / / I'Jrt-tMrE ptt4E )

-.

5;l'-1"..*, .({ir r . .+.,e

tlJ t.F.'E 5r,{RE
-
. q,*Ls ..L)
,.6,

ü€9..*."
1+ * Ad
-r

Section 1: The Deueloqnent of MS DOS

 1983-198,1

The |ntel80286 micra-

processar, the chip at
the heart o.f the lBM
PC/AT. uhrch is shobn
beside it. Version 3.A of
MS DOS, deueloped.fol
this machine, offered
suppartfor netuofks
an.l the PC/AT'S 1.2
negabtte ftappy disk
dl.iue 6ntl builtin
CMOS clack.

But version 3.0 was net a simple extension of version 2.0.

In iaying the f.oundation for net_
working, the MS-DOS ream had complerely redesignecl and rewriiten
the DOS kernel.
Dilferent as it was from version 1.0, version 2.0 had been built on
top ofthe same structure.
For example, whcreas file requests in MS DOS 1.0 used FCBs,
requests in version 2.0 used
file handles. However, the version 2.0 handle calls woultl simply parse
the pathname and
then use the underlying FCB calls in the same way as version 1.'0.
The redirected input and
output in yersion 2.0 further compiicated the file-system requests. Vhen
a progra; used
one ofthe CP/M-compatible calls for character input or output, MS-DOS
2.0 first opened a
handle and then turned it back into an FCB call at a lower level. Version
3.0 eliminated this
redundancy by eliminating the old FCB input/output code ofvemions
1 and 2, replacing it
with a standard set of L/O calls thar could be callecl directly by both
FCts ca s and handle
calls The look alike carls for cp/M compatibre character Iio were included
as part ofthe
scr of handle calls. As a result of this restructuring, these
calls were distinctlv faster in
version 3.0 than in version 2.0.

More impoftant than the elimination of inefficiencies, however, was

the fact that this new
structure made it easier to handle network requests under the ISO
Open System Intercon_
nect model Microsoft was using for networking. The ISO model
describes a number of
protocol layers, ranging from the application_to_application interface
at the top level down
to the physical link-plugging inro the network-at rhe lowest
level. In the middle is the
transport laye1 which manages the actual transfer of data. The layers
above the transport
layer belong to thc realn ofthe operating system; the layers
below the transport layer are
traditionally the domxin ofthe nctwork software or hardware.
On the IBM PC network, the transport layer and the server lunctions
were handled bv
IBM s Network Adapter card ancl the task of MS DOS was to support
this hardware. FÄr its
other OEM customers, howeveq Microsoft neecled to supply boÄ the
transport ancl the
serler functions as software. Although version 3.0 clid not provide this general_purpose
net\.orking software, it did provide the basic support for IBM,s networking
haräware.
The support lor IBM consisted ()f redirector and sharer software.
MS DOS used an ap-
proach to netllrrrking in whiclt remote rcquests were routed
by a redirector that was able

=: The .\IS-D?S Encjctopertia

 1981

was composed of
to interact with the transport layer of the network The transport layer
the device drivers that could reiiably transfer data from one part of the network to another
newly designed low-level file VO code' the operating sys-
Just before a call was sent to the
tem determined whether the call was local & remote A local call
would be allowed to fall
which'
through to the local file I/O co<te; a remote call would be passed to the redirector
would make the resources on a remote machine
work;g with the of,erating system,
appear as if theY were local.

Version 3.L
card were in place
Both the redirector and the sharer interfaces for IBM's Network Adapter
itself wasn't ready Version
in version 3.0 when it was delivered to IBM, but the redirector
anti Reynolds and released th'ee months latel completed this
3.1, completed by Zbikowski
..i*ork atppoti made it available in the form of Microsoft Networks for use on non
"nd
IBM network cards.
" Services
Microsoft Networks was built on the concept of "services" and "consumers
application and ran on a
were provided by a file server, which was part ofthe Networks
programs on various network machines'
cornputer dedicated to the task Consumeß were
file server; it was then the
Requests for information were passed at a high levet to the
information on the disk.
responsibitity ofthe file serverio determine where to find the
need knowledge of the remote
The requesting programs-the consumers did not any
machine, not even what type offile system it had'
having to know the
This ability to pass a highJevel requesr to a remote server without
allowed another level of generalization ofthe system
detaits of the server's file structur;
file systems could be accessed on the same network lt
In MS-DOS 3.1, different types of
was possible, for ex"-pl.,io u XEND( machine across the network from an
"tctl"
data from XEND( files'
MS-DOS machine and to read
I
Yet the variability of the
Microsoft Networks was designed to be hardware independent
was maior problem in developing
classes of programs that would be using its structures
a
In evaluating this variability'
a networking-"system that would be transparent to the user'
Microsoft identified three types of programs:
a used only the documented
First were the MS-DOS-compatible programs These
requesting services from the operating system and
software-interrupt rnethod äf
would run on any MS DOS machine without problems'
a Second were the MS-DOS based programs These
would run on lBM-compatible
computers but not necessarily on all MS DOS machines'
. of MS-DOS orthat
Third were the programs th;used undocumentecl features
directly These programs tended to have the best perfor-
adclressed the hardware
mance but were also the most difficult to support

of rhese. Microsoft officially encouraged the wfiting of

Ms-Dos-compatible programs for
use on the network

Sectian I: The Deuelapnent ,JMS DOS 43

 19n6

Network concerns
The file-access module was changed in vc.sion 3.0 to simplifv fiie management on the
network' but this did not solve arl the probrems. For instance. \rS-DoS still needed to han-
dle FCB requests from programs thar used rhem. but manr- programs *,ould open an FCB
and never close it. one ofthe functions ofrhe server q'as to keep track ofall opren files
on the network, and it ran into difficurties s'hen an FcB was opened 50 or 100iimes ancl
never closed. To solve this problem, Microsoft introduced an FCB cache in version 3.1 that
allowed only four FCBs to be open at any one time. Ifa fifth FCB was opened, the least re,
cently used one was closed automatically and released. In addition, an FCBS command
was added in the CONFIG.SYS file to allow the user or network manager to change the
maximum number ofFCBs that could be open at any one time and to protect some ofthe
FCBs from automatic closure.

In general, the logical device independence that had been a goal of MS_DOS acquired new
meaning-and generated new problems-with networking. One problem concerned
printers on the network. Commonly, networks are used to allow several people to share a
pdnter. The netü,ork could easily accommodate a program that woulcl open the printer,
write to it, and close it again. Some programs, howevel woulcl try to use the dire;t IBM
BIOS interface to access the printer. To handle this situation, Microsoft,s designers had to
develop a way for MS DOS to intercept these BIOS requests and filter out the ones the
server could not handle. Once this was accomplished, version 3.1was able to hanclle most
types of printer output on the network in a transparent manner.

Version J.2
InJanuary 1986, Microsofr released another revision of MS-DOS, version 1.2, which
supported 372-inch floppy disks. Version 3.2 also moved the formarting funcrion for a
device out ofthe FORMAT utility routine and into the device driver, eliminating the need
for a special hardwaredependent program in addition to the device driver. It included a
sample installable block-device driver and, finally, benefited the users and manufacturers
of IBM-compatible computers by including maior rewrires of the MS-DOS utilities to
increase compatibiliry with those of IBM.

14 The MS DOS Encyclopetlia

 1987

The Future

Since its appearance in 1981, MS-DOS has taken and held an enviable position in the
microcomputer environment. Not only has it "taught" millions of personal computers
"ho$/ to think," it has taught equal millions of people how to use computers Many highly
sophisticated computer users can trace their first encounter with these machines to the
oiginal IBM PC and version 1.0 of MS-DOS. The MS-DOS command interface is the one
with which they are comfortable and it is the MS-DOS file structure that, in one way or
anothet they wander through with familiarity
Microsoft has stated its commitment to ensuring that. for the foreseeable future, MS-DOS
will continue to evolve and grov, changing as it has done in the past to satisfy the needs of
its millions of users. In the long term, MS DoS, the product of a surprisingly small group of
gifted people, will undoubtedly remain the industry standard for as long as 8086-based
(and to some extent, 80286-based) microcomputers exist in the business world The story
of MS-DOS will, of course, remain even longer' For this operating system has earned its
place in microcomputing history.

JoAnrLe Woodcoch

section I: The DetElopment of MS-Dos 45