Documente Academic
Documente Profesional
Documente Cultură
AppendixATheTanenbaumTorvaldsDebate
OpenSources:VoicesfromtheOpenSourceRevolution
1stEditionJanuary1999
1565925823,OrderNumber:5823
280pages,$24.95
AppendixA
TheTanenbaumTorvaldsDebate
WhatfollowsinthisappendixarewhatareknowninthecommunityastheTanenbaum/Linus"Linuxisobsolete"debates.Andrew
TanenbaumisawellrespectedresearcherwhohasmadeaverygoodlivingthinkingaboutoperatingsystemsandOSdesign.Inearly
1992,noticingthewaythattheLinuxdiscussionhadtakenoverthediscussionincomp.os.minix,hedecideditwastimetocommenton
Linux.
AlthoughAndrewTanenbaumhasbeenderidedforhisheavyhandandmisjudgementsoftheLinuxkernel,suchareactiontoTanenbaum
isunfair.WhenLinushimselfheardthatwewereincludingthis,hewantedtomakesurethattheworldunderstoodthatheholdsno
animustowardsTanenbaumandinfactwouldnothavesanctioneditsinclusionifwehadnotbeenabletoconvincehimthatitwould
showthewaytheworldwasthinkingaboutOSdesignatthetime.
WefelttheinclusionofthisappendixwouldgiveagoodperspectiveonhowthingswerewhenLinuswasunderpressurebecausehe
abandonedtheideaofmicrokernelsinacademia.ThefirstthirdofLinus'essaydiscussesthisfurther.
ElectroniccopiesofthisdebateareavailableontheWebandareeasilyfoundthroughanysearchservice.It'sfuntoreadthisandnote
whojoinedintothediscussionyouseeuserhackerKenThompson(oneofthefoundersofUnix)andDavidMiller(whoisamajor
Linuxkernelhackernow),aswellasmanyothers.
Toputthisdiscussionintoperspective,whenitoccurredin1992,the386wasthedominatingchipandthe486hadnotcomeoutonthe
market.MicrosoftwasstillasmallcompanysellingDOSandWordforDOS.Lotus123ruledthespreadsheetspaceandWordPerfectthe
wordprocessingmarket.DBASEwasthedominantdatabasevendorandmanycompaniesthatarehouseholdnamestodayNetscape,
Yahoo,Excitesimplydidnotexist.
From:ast@cs.vu.nl(AndyTanenbaum)
Newsgroups:comp.os.minix
Subject:LINUXisobsolete
Date:29Jan9212:12:50GMT
IwasintheU.S.foracoupleofweeks,soIhaven'tcommentedmuchon
LINUX(notthatIwouldhavesaidmuchhadIbeenaround),butforwhat
itisworth,Ihaveacoupleofcommentsnow.
Asmostofyouknow,formeMINIXisahobby,somethingthatIdointhe
eveningwhenIgetboredwritingbooksandtherearenomajorwars,
revolutions,orsenatehearingsbeingtelevisedliveonCNN.Myreal
jobisaprofessorandresearcherintheareaofoperatingsystems.
Asaresultofmyoccupation,IthinkIknowabitaboutwhereoperating
aregoinginthenextdecadeorso.Twoaspectsstandout:
1.MICROKERNELVSMONOLITHICSYSTEM
Mostolderoperatingsystemsaremonolithic,thatis,thewholeoperating
systemisasinglea.outfilethatrunsin'kernelmode.'Thisbinary
containstheprocessmanagement,memorymanagement,filesystemandthe
rest.ExamplesofsuchsystemsareUNIX,MSDOS,VMS,MVS,OS/360,
MULTICS,andmanymore.
Thealternativeisamicrokernelbasedsystem,inwhichmostoftheOS
runsasseparateprocesses,mostlyoutsidethekernel.Theycommunicate
bymessagepassing.Thekernel'sjobistohandlethemessagepassing,
interrupthandling,lowlevelprocessmanagement,andpossiblytheI/O.
ExamplesofthisdesignaretheRC4000,Amoeba,Chorus,Mach,andthe
notyetreleasedWindows/NT.
WhileIcouldgointoalongstoryhereabouttherelativemeritsofthe
twodesigns,sufficeittosaythatamongthepeoplewhoactuallydesign
operatingsystems,thedebateisessentiallyover.Microkernelshavewon.
Theonlyrealargumentformonolithicsystemswasperformance,andthere
isnowenoughevidenceshowingthatmicrokernelsystemscanbejustas
http://www.oreilly.com/openbook/opensources/book/appa.html
1/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
fastasmonolithicsystems(e.g.,RickRashidhaspublishedpaperscomparing
Mach3.0tomonolithicsystems)thatitisnowalloverbuttheshoutin'.
MINIXisamicrokernelbasedsystem.Thefilesystemandmemorymanagement
areseparateprocesses,runningoutsidethekernel.TheI/Odriversare
alsoseparateprocesses(inthekernel,butonlybecausethebraindead
natureoftheIntelCPUsmakesthatdifficulttodootherwise).LINUXis
amonolithicstylesystem.Thisisagiantstepbackintothe1970s.
Thatisliketakinganexisting,workingCprogramandrewritingitin
BASIC.Tome,writingamonolithicsystemin1991isatrulypooridea.
2.PORTABILITY
Onceuponatimetherewasthe4004CPU.Whenitgrewupitbecamean
8008.Thenitunderwentplasticsurgeryandbecamethe8080.Itbegat
the8086,whichbegatthe8088,whichbegatthe80286,whichbegatthe
80386,whichbegatthe80486,andsoonuntotheNthgeneration.In
themeantime,RISCchipshappened,andsomeofthemarerunningatover
100MIPS.Speedsof200MIPSandmorearelikelyinthecomingyears.
Thesethingsarenotgoingtosuddenlyvanish.Whatisgoingtohappen
isthattheywillgraduallytakeoverfromthe80x86line.Theywill
runoldMSDOSprogramsbyinterpretingthe80386insoftware.(Ieven
wrotemyownIBMPCsimulatorinC,whichyoucangetbyFTPfrom
ftp.cs.vu.nl=192.31.231.42indirminix/simulator.)Ithinkitisa
grosserrortodesignanOSforanyspecificarchitecture,sincethatis
notgoingtobearoundallthatlong.
MINIXwasdesignedtobereasonablyportable,andhasbeenportedfromthe
Intellinetothe680x0(Atari,Amiga,Macintosh),SPARC,andNS32016.
LINUXistiedfairlycloselytothe80x86.Notthewaytogo.
Don'tgetmewrong,IamnotunhappywithLINUX.Itwillgetallthepeople
whowanttoturnMINIXinBSDUNIXoffmyback.Butinallhonesty,Iwould
suggestthatpeoplewhowanta**MODERN**"free"OSlookaroundfora
microkernelbased,portableOS,likemaybeGNUorsomethinglikethat.
AndyTanenbaum(ast@cs.vu.nl)
P.S.Justasarandomaside,AmoebahasaUNIXemulator(runninginuser
space),butitisfarfromcomplete.Ifthereareanypeoplewhowould
liketoworkonthat,pleaseletmeknow.TorunAmoebayouneedafew386s,
oneofwhichneeds16M,andallofwhichneedtheWDEthernetcard.
From:torvalds@klaava.Helsinki.FI(LinusBenedictTorvalds)
Subject:Re:LINUXisobsolete
Date:29Jan9223:14:26GMT
Organization:UniversityofHelsinki
Well,withasubjectlikethis,I'mafraidI'llhavetoreply.
Apologiestominixuserswhohaveheardenoughaboutlinuxanyway.I'd
liketobeabletojust"ignorethebait",but...Timeforsome
seriousflamefesting!
Inarticle<12595@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>
>IwasintheU.S.foracoupleofweeks,soIhaven'tcommentedmuchon
>LINUX(notthatIwouldhavesaidmuchhadIbeenaround),butforwhat
>itisworth,Ihaveacoupleofcommentsnow.
>
>Asmostofyouknow,formeMINIXisahobby,somethingthatIdointhe
>eveningwhenIgetboredwritingbooksandtherearenomajorwars,
>revolutions,orsenatehearingsbeingtelevisedliveonCNN.Myreal
>jobisaprofessorandresearcherintheareaofoperatingsystems.
Youusethisasanexcuseforthelimitationsofminix?Sorry,butyou
loose:I'vegotmoreexcusesthanyouhave,andlinuxstillbeatsthe
pantsofminixinalmostallareas.Nottomentionthefactthatmost
ofthegoodcodeforPCminixseemstohavebeenwrittenbyBruceEvans.
Re1:youdoingminixasahobbylookatwhomakesmoneyoffminix,
andwhogiveslinuxoutforfree.Thentalkabouthobbies.Makeminix
freelyavailable,andoneofmybiggestgripeswithitwilldisappear.
Linuxhasverymuchbeenahobby(butaseriousone:thebesttype)for
me:Igetnomoneyforit,andit'snotevenpartofanyofmystudies
intheuniversity.I'vedoneitallonmyowntime,andonmyown
machine.
Re2:yourjobisbeingaprofessorandresearcher:That'sonehellofa
goodexcuseforsomeofthebraindamagesofminix.Icanonlyhope(and
assume)thatAmoebadoesn'tsucklikeminixdoes.
>1.MICROKERNELVSMONOLITHICSYSTEM
True,linuxismonolithic,andIagreethatmicrokernelsarenicer.With
alessargumentativesubject,I'dprobablyhaveagreedwithmostofwhat
yousaid.Fromatheoretical(andaesthetical)standpointlinuxlooses.
IftheGNUkernelhadbeenreadylastspring,I'dnothavebotheredto
evenstartmyproject:thefactisthatitwasn'tandstillisn't.Linux
winsheavilyonpointsofbeingavailablenow.
>MINIXisamicrokernelbasedsystem.[deleted,butnotsothatyou
>missthepoint]LINUXisamonolithicstylesystem.
http://www.oreilly.com/openbook/opensources/book/appa.html
2/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
Ifthiswastheonlycriterionforthe"goodness"ofakernel,you'dbe
right.Whatyoudon'tmentionisthatminixdoesn'tdothemicrokernel
thingverywell,andhasproblemswithrealmultitasking(inthe
kernel).IfIhadmadeanOSthathadproblemswithamultithreading
filesystem,Iwouldn'tbesofasttocondemnothers:infact,I'ddomy
damndesttomakeothersforgetaboutthefiasco.
[yes,Iknowtherearemultithreadinghacksforminix,buttheyare
hacks,andbruceevanstellsmetherearelotsofraceconditions]
>2.PORTABILITY
"Portabilityisforpeoplewhocannotwritenewprograms"
me,rightnow(withtongueincheek)
Thefactisthatlinuxismoreportablethanminix.What?Ihearyou
say.It'struebutnotinthesensethatastmeans:Imadelinuxas
conformanttostandardsasIknewhow(withouthavinganyPOSIXstandard
infrontofme).Portingthingstolinuxisgenerally/much/easier
thanportingthemtominix.
Iagreethatportabilityisagoodthing:butonlywhereitactuallyhas
somemeaning.Thereisnoideaintryingtomakeanoperatingsystem
overlyportable:adheringtoaportableAPIisgoodenough.Thevery
/idea/ofanoperatingsystemistousethehardwarefeatures,andhide
thembehindalayerofhighlevelcalls.Thatisexactlywhatlinux
does:itjustusesabiggersubsetofthe386featuresthanother
kernelsseemtodo.Ofcoursethismakesthekernelproperunportable,
butitalsomakesfora/much/simplerdesign.Anacceptabletradeoff,
andonethatmadelinuxpossibleinthefirstplace.
Ialsoagreethatlinuxtakesthenonportabilitytoanextreme:Igot
my386lastJanuary,andlinuxwaspartlyaprojecttoteachmeabout
it.Manythingsshouldhavebeendonemoreportablyifitwouldhave
beenarealproject.I'mnotmakingoverlymanyexcusesaboutit
though:itwasadesigndecision,andlastaprilwhenIstartedthe
thing,Ididn'tthinkanybodywouldactuallywanttouseit.I'mhappy
toreportIwaswrong,andasmysourceisfreelyavailable,anybodyis
freetotrytoportit,eventhoughitwon'tbeeasy.
Linus
PS.Iapologiseforsometimessoundingtooharsh:minixisniceenough
ifyouhavenothingelse.Amoebamightbeniceifyouhave510spare
386'slyingaround,butIcertainlydon't.Idon'tusuallygetinto
flames,butI'mtouchywhenitcomestolinux:)
From:ast@cs.vu.nl(AndyTanenbaum)
Subject:Re:LINUXisobsolete
Date:30Jan9213:44:34GMT
Inarticle<1992Jan29.231426.20469@klaava.Helsinki.FI>torvalds@klaava.Helsinki.
FI(LinusBenedictTorvalds)writes:
>Youusethis[beingaprofessor]asanexcuseforthelimitationsofminix?
ThelimitationsofMINIXrelateatleastpartlytomybeingaprofessor:
Anexplicitdesigngoalwastomakeitrunoncheaphardwaresostudents
couldaffordit.Inparticular,foryearsitranonaregular4.77MHZPC
withnoharddisk.Youcoulddoeverythinghereincludingmodifyandrecompile
thesystem.Justfortherecord,asofabout1yearago,thereweretwo
versions,oneforthePC(360Kdiskettes)andoneforthe286/386(1.2M).
ThePCversionwasoutsellingthe286/386versionby2to1.Idon'thave
figures,butmyguessisthatthefractionofthe60millionexistingPCsthat
are386/486machinesasopposedto8088/286/680x0etcissmall.Amongstudents
itisevensmaller.Makingsoftwarefree,butonlyforfolkswithenoughmoney
tobuyfirstclasshardwareisaninterestingconcept.
Ofcourse5yearsfromnowthatwillbedifferent,but5yearsfromnow
everyonewillberunningfreeGNUontheir200MIPS,64MSPARCstation5.
>Re2:yourjobisbeingaprofessorandresearcher:That'sonehellofa
>goodexcuseforsomeofthebraindamagesofminix.Icanonlyhope(and
>assume)thatAmoebadoesn'tsucklikeminixdoes.
Amoebawasnotdesignedtorunonan8088withnoharddisk.
>Ifthiswastheonlycriterionforthe"goodness"ofakernel,you'dbe
>right.Whatyoudon'tmentionisthatminixdoesn'tdothemicrokernel
>thingverywell,andhasproblemswithrealmultitasking(inthe
>kernel).IfIhadmadeanOSthathadproblemswithamultithreading
>filesystem,Iwouldn'tbesofasttocondemnothers:infact,I'ddomy
>damndesttomakeothersforgetaboutthefiasco.
Amultithreadedfilesystemisonlyaperformancehack.Whenthereisonly
onejobactive,thenormalcaseonasmallPC,itbuysyounothingandadds
complexitytothecode.Onmachinesfastenoughtosupportmultipleusers,
youprobablyhaveenoughbuffercachetoinsureahitcachehitrate,in
whichcasemultithreadingalsobuysyounothing.Itisonlyawinwhenthere
aremultipleprocessesactuallydoingrealdiskI/O.Whetheritisworth
makingthesystemmorecomplicatedforthiscaseisatleastdebatable.
Istillmaintainthepointthatdesigningamonolithickernelin1991is
afundamentalerror.Bethankfulyouarenotmystudent.Youwouldnot
getahighgradeforsuchadesign:)
>Thefactisthatlinuxismoreportablethanminix.What?Ihearyou
http://www.oreilly.com/openbook/opensources/book/appa.html
3/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
>say.It'struebutnotinthesensethatastmeans:Imadelinuxas
>conformanttostandardsasIknewhow(withouthavinganyPOSIXstandard
>infrontofme).Portingthingstolinuxisgenerally/much/easier
>thanportingthemtominix.
MINIXwasdesignedbeforePOSIX,andisnowbeing(slowly)POSIXizedas
everyonewhofollowsthisnewsgroupknows.Everyoneagreesthatuserlevel
standardsareagoodidea.Asanaside,Icongratulateyouforbeingable
towriteaPOSIXconformantsystemwithouthavingthePOSIXstandardinfront
ofyou.Ifinditdifficultenoughafterstudyingthestandardatgreatlength.
Mypointisthatwritinganewoperatingsystemthatiscloselytiedtoany
particularpieceofhardware,especiallyaweirdoneliketheIntelline,
isbasicallywrong.AnOSitselfshouldbeeasilyportabletonewhardware
platforms.WhenOS/360waswritteninassemblerfortheIBM360
25yearsago,theyprobablycouldbeexcused.WhenMSDOSwaswritten
specificallyforthe8088tenyearsago,thiswaslessthanbrilliant,as
IBMandMicrosoftnowonlytoopainfullyrealize.WritinganewOSonlyforthe
386in1991getsyouyoursecond'F'forthisterm.Butifyoudorealwell
onthefinalexam,youcanstillpassthecourse.
Prof.AndrewS.Tanenbaum(ast@cs.vu.nl)
From:feustel@netcom.COM(DavidFeustel)
Subject:Re:LINUXisobsolete
Date:30Jan9218:57:28GMT
Organization:DAFCOAnOS/2Oasis
ast@cs.vu.nl(AndyTanenbaum)writes:
>Istillmaintainthepointthatdesigningamonolithickernelin1991is
>afundamentalerror.Bethankfulyouarenotmystudent.Youwouldnot
>getahighgradeforsuchadesign:)
That'sok.Einsteingotlousygradesinmathandphysics.
From:pete@ohm.york.ac.uk(PeteFrench.)
Subject:Re:LINUXisobsolete
Date:31Jan9209:49:37GMT
Organization:ElectronicsDepartment,UniversityofYork,UK
inarticle<1992Jan30.195850.7023@epas.toronto.edu>,meggin@epas.utoronto.ca
(DavidMegginson)says:
>
>Inarticle<1992Jan30.185728.26477feustel@netcom.COM>feustel@netcom.COM
(David>Feustel)writes:
>>
>>That'sok.Einsteingotlousygradesinmathandphysics.
>
>AndDanQuaylegotlowgradesinpoliticalscience.Ithinkthatthere
>aremoreDanQuaylesthanEinsteinsoutthere...;)
Whatahorriblethought!
Butonthepointsaboutmicrokernelvmonolithic,isntthispartlyan
artifactofthelanguagebeingused?MINIXmaywellbedesignedasa
microkernelsystem,butintheendyoustillendupwithalarge
monolithicchunkofbinarydatathatgetsloadedinas"theOS".Isntit
writtenasseparateprogramssimplybecauseCdoesnotsupporttheidea
ofmultipleprocesseswithinasinglepieceofmonolithiccode.Isthere
anyrealdifferencebetweenamicrokernelwrittenasseveralpiecesofC
andamonolithickernelwritteninsomethinglikeOCCAM?Iwouldhave
thoughtthatinthiscasethemonolithicdesignwouldbeabetterone
thanthemicorkernelstylesincewiththeadvantageofinbuilt
languageconcurrencythekernelcouldbemadeevenmoremodularthanthe
MINIXoneis.
AnyoneforMINOX:)
bat.
From:kt4@prism.gatech.EDU(KenThompson)
Subject:Re:LINUXisobsolete
Date:3Feb9223:07:54GMT
Organization:GeorgiaInstituteofTechnology
viewpointmaybelargelyunrelatedtoitsusefulness.Manyifnot
mostofthesoftwareweuseisprobablyobsoleteaccordingtothe
latestdesigncriteria.Mostuserscouldprobablycarelessifthe
internalsoftheoperatingsystemtheyuseisobsolete.Theyare
rightlymoreinterestedinitsperformanceandcapabilitiesatthe
userlevel.
Iwouldgenerallyagreethatmicrokernelsareprobablythewaveof
thefuture.However,itisinmyopinioneasiertoimplementa
monolithickernel.Itisalsoeasierforittoturnintoamessin
ahurryasitismodified.
Regards,
http://www.oreilly.com/openbook/opensources/book/appa.html
4/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
Ken
From:kevin@taronga.taronga.com(KevinBrown)
Subject:Re:LINUXisobsolete
Date:4Feb9208:08:42GMT
Organization:UniversityofHouston
Inarticle<47607@hydra.gatech.EDU>kt4@prism.gatech.EDU(KenThompson)writes:
>viewpointmaybelargelyunrelatedtoitsusefulness.Manyifnot
>mostofthesoftwareweuseisprobablyobsoleteaccordingtothe
>latestdesigncriteria.Mostuserscouldprobablycarelessifthe
>internalsoftheoperatingsystemtheyuseisobsolete.Theyare
>rightlymoreinterestedinitsperformanceandcapabilitiesatthe
>userlevel.
>
>Iwouldgenerallyagreethatmicrokernelsareprobablythewaveof
>thefuture.However,itisinmyopinioneasiertoimplementa
>monolithickernel.Itisalsoeasierforittoturnintoamessin
>ahurryasitismodified.
Howdifficultisittostructurethesourcetreeofamonolithickernel
suchthatmostmodificationsdon'thavealargenegativeimpactonthe
source?Whatsortsofpitfallsdoyourunintointhissortofendeavor,
andwhatsuggestionsdoyouhavefordealingwiththem?
IguesswhatI'maskingis:howdifficultisittoorganizethesource
suchthatmostchangestothekernelremainlocalizedinscope,even
thoughthekernelitselfismonolithic?
Ifigureyou'vegotyearsofexperiencewithmonolithickernels:),
soI'dthinkyou'dhavethebestshotatansweringquestionslike
these.
KevinBrown
From:rburns@finess.Corp.Sun.COM(RandyBurns)
Subject:Re:LINUXisobsolete
Date:30Jan9220:33:07GMT
Organization:SunMicrosystems,Mt.View,Ca.
Inarticle<12615@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>Inarticle<1992Jan29.231426.20469@klaava.Helsinki.FI>torvalds@klaava.Helsinki.
>FI(LinusBenedictTorvalds)writes:
>Ofcourse5yearsfromnowthatwillbedifferent,but5yearsfromnow
>everyonewillberunningfreeGNUontheir200MIPS,64MSPARCstation5.
Well,Iforonewould_love_toseethishappen.
>>Thefactisthatlinuxismoreportablethanminix.What?Ihearyou
>>say.It'struebutnotinthesensethatastmeans:Imadelinuxas
>>conformanttostandardsasIknewhow(withouthavinganyPOSIXstandard
>>infrontofme).Portingthingstolinuxisgenerally/much/easier
>>thanportingthemtominix.
.........
>Mypointisthatwritinganewoperatingsystemthatiscloselytiedtoany
>particularpieceofhardware,especiallyaweirdoneliketheIntelline,
>isbasicallywrong.
Firstoff,thepartsofLinuxtunedmostfinelytothe80x86aretheKernel
andthedevices.MyownsenseisthatevenifLinuxissimplyastopgap
measuretoletusallrunGNUsoftware,itisstillworthwhiletohavea
afinelytunedkernelforthemostnumerousarchitecturepresentlyin
existance.
>AnOSitselfshouldbeeasilyportabletonewhardware
>platforms.
Well,theonlypartofLinuxthatisn'tportableisthekernelanddrivers.
Comparetothecompilers,utilities,windowingsystemetc.thisisreally
asmallpartoftheeffort.SinceLinuxhasalargedegreeofcall
compatibilitywithportableOS'sIwouldn'tcomplain.I'mpersonally
verygratefultohaveanOSthatmakesitmorelikelythatsomeofuswill
beabletotakeadvantageofthesoftwarethathascomeoutofBerkeley,
FSF,CMUetc.Itmaywellbethatin23yearswhenultracheapBSD
variantsandHurdproliferate,thatLinuxwillbeobsolete.Still,right
nowLinuxgreatlyreducesthecostofusingtoolslikegcc,bison,bash
whichareusefulinthedevelopmentofsuchanOS.
From:torvalds@klaava.Helsinki.FI(LinusBenedictTorvalds)
Subject:Re:LINUXisobsolete
Date:31Jan9210:33:23GMT
Organization:UniversityofHelsinki
Inarticle<12615@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>ThelimitationsofMINIXrelateatleastpartlytomybeingaprofessor:
>Anexplicitdesigngoalwastomakeitrunoncheaphardwaresostudents
>couldaffordit.
Allright:arealtechnicalpoint,andonethatmadesomeofmycomments
inexcusable.Butatthesametimeyoushootyourselfinthefootabit:
http://www.oreilly.com/openbook/opensources/book/appa.html
5/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
nowyouadmitthatsomeoftheerrorsofminixwerethatitwastoo
portable:includingmachinesthatweren'treallydesignedtorununix.
Thatassumptionleadtothefactthatminixnowcannoteasilybe
extendedtohavethingslikepaging,evenformachinesthatwould
supportit.Yes,minixisportable,butyoucanrewritethatas
"doesn'tuseanyfeatures",andstillberight.
>Amultithreadedfilesystemisonlyaperformancehack.
Nottrue.It'saperformancehack/onamicrokernel/,butit'san
automaticfeaturewhenyouwriteamonolithickerneloneareawhere
microkernelsdon'tworktoowell(asIpointedoutinmypersonalmail
toast).Whenwritingaunixthe"obsolete"way,youautomaticallyget
amultithreadedkernel:everyprocessdoesit'sownjob,andyoudon't
havetomakeuglythingslikemessagequeuestomakeitwork
efficiently.
Besides,therearepeoplewhowouldconsider"onlyaperformancehack"
vital:unlessyouhaveacray3,I'dguesseverybodygetstiredof
waitingonthecomputerallthetime.IknowIdidwithminix(andyes,
Idowithlinuxtoo,butit's/much/better).
>Istillmaintainthepointthatdesigningamonolithickernelin1991is
>afundamentalerror.Bethankfulyouarenotmystudent.Youwouldnot
>getahighgradeforsuchadesign:)
Well,Iprobablywon'tgettoogoodgradesevenwithoutyou:Ihadan
argument(completelyunrelatednotevenpertainingtoOS's)withthe
personhereattheuniversitythatteachesOSdesign.Iwonderwhen
I'lllearn:)
>Mypointisthatwritinganewoperatingsystemthatiscloselytiedtoany
>particularpieceofhardware,especiallyaweirdoneliketheIntelline,
>isbasicallywrong.
But/my/pointisthattheoperatingsystem/isn't/tiedtoany
processorline:UNIXrunsonmostrealprocessorsinexistence.Yes,
the/implementation/ishardwarespecific,butthere'saHUGE
difference.YoumentionOS/360andMSDOGasexamplesofbaddesigns
astheywerehardwaredependent,andIagree.Butthere'sabig
differencebetweentheseandlinux:linuxAPIisportable(notduetomy
cleverdesign,butduetothefactthatIdecidedtogoforafairly
wellthoughtoutandtestedOS:unix.)
Ifyouwriteprogramsforlinuxtoday,youshouldn'thavetoomany
surpriseswhenyoujustrecompilethemforHurdinthe21stcentury.As
hasbeennoted(notonlybyme),thelinuxkernelisaminisculepartof
acompletesystem:Fullsourcesforlinuxcurrentlyrunstoabout200kB
compressedfullsourcestoasomewhatcompletedevelopementsystemis
atleast10MBcompressed(andeasilymuch,muchmore).Andallofthat
sourceisportable,exceptforthistinykernelthatyoucan(provably:
Ididit)rewritetotallyfromscratchinlessthanayearwithout
having/any/priorknowledge.
Infactthe/whole/linuxkernelismuchsmallerthanthe386dependent
thingsinmach:i386.tar.Zforthecurrentversionofmachiswellover
800kBcompressed(823391bytesaccordingtonic.funet.fi).Admittedly,
machis"somewhat"biggerandhasmorefeatures,butthatshouldstill
tellyousomething.
Linus
From:kaufman@eecs.nwu.edu(MichaelL.Kaufman)
Subject:Re:LINUXisobsolete
Date:3Feb9222:27:48GMT
Organization:EECSDepartment,NorthwesternUniversity
Itriedtosendthesetwopostsfromwork,butIthinktheygoteaten.Ifyou
haveseenthemalready,sorry.
AndyTanenbaumwritesaninterestingarticle(alsointerestingwasfindingout
thatheactuallyreadsthisgroup)butIthinkheismissinganimportant
point.
HeWrote:
>Asmostofyouknow,formeMINIXisahobby,...
Whichisalsoprobablytrueofmost,ifnotall,ofthepeoplewhoareinvolved
inLinux.WearenotdevelopingasystemtotakeovertheOSmarket,weare
justhavingagoodtime.
>Whatisgoingtohappen
>isthattheywillgraduallytakeoverfromthe80x86line.Theywill
>runoldMSDOSprogramsbyinterpretingthe80386insoftware.
Wellwhenthishappens,ifIstillwanttoplaywithLinux,Icanjustrunit
onmy386simulator.
>MINIXwasdesignedtobereasonablyportable,andhasbeenportedfromthe
>Intellinetothe680x0(Atari,Amiga,Macintosh),SPARC,andNS32016.
http://www.oreilly.com/openbook/opensources/book/appa.html
6/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
>LINUXistiedfairlycloselytothe80x86.Notthewaytogo.
That'sfineforthepeoplewhohavethosemachines,butitwasn'tafree
lunch.Thatportibilitywasgainedatthecostofsomeperformanceandsome
featuresonthe386.BeforeyoudecidethatLINUXisnotthewaytogo,you
shouldthinkaboutwhatitisgoingtobeusedfor.Iamgoingtouseitfor
runningmemoryandcomputationintensivegraphicsprogramsonmy486.Forme,
speedandmemoryweremoreimportantthenfuturestateoftheartnessand
portability.
>Butinallhonesty,Iwould
>suggestthatpeoplewhowanta**MODERN**"free"OSlookaroundfora
>microkernelbased,portableOS,likemaybeGNUorsomethinglikethat.
Idon'tknowofanyfreemicrokernelbased,portableOSes.GNUisstill
vaporware,andlikelytoremainthatwayfortheforseeablefuture.Do
youactuallyhaveonetorecomend,orareyoujusttoyingwithme?;)
Inarticle<12615@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>Mypointisthatwritinganewoperatingsystemthatiscloselytiedtoany
>particularpieceofhardware,especiallyaweirdoneliketheIntelline,
>isbasicallywrong.AnOSitselfshouldbeeasilyportabletonewhardware
>platforms.
IthinkIseewhereIdisagreewithyounow.YouarelookingatOSdesign
asanendinitself.Minixisgoodbecauseitisportable/MicroKernal/etc.
Linuxisnotgoodbecauseitismonolithic/tightlytiedtoIntel/etc.That
isnotastrangeattitudeforsomeoneintheacedemicworld,butitisnot
somethingyoushouldexpecttobeuniversallyshared.Linuxisnotbeingwritten
asateachingtool,orasanabstractexercise.Itisbeingwrittentoallow
peopletorunGNUtypesoftware_today_.Thefactthatitmaynotbeinuse
infiveyearsislessimportantthenthefactthattoday(well,byApril
probably)IcanrunallsortsofsoftwareonitthatIwanttorun.Youkeep
sayingthatMinixisbetter,butifitwillnotrunthesoftwarethatIwant
torun,itreallyisn'tthatgood(forme)atall.
>WhenOS/360waswritteninassemblerfortheIBM360
>25yearsago,theyprobablycouldbeexcused.WhenMSDOSwaswritten
>specificallyforthe8088tenyearsago,thiswaslessthanbrilliant,as
>IBMandMicrosoftnowonlytoopainfullyrealize.
Samepoint.MSoftdidnotcomeoutwithDosto"explorethefrontiersofos
research".Theydidittomakeabuck.AndconsideringthefactthatMSDOS
probablystilloutsellseveryoneelseputtogether,Idon'tthinkthatyou
saythattheyhavefailed_intheirgoals_.NotthatMSDosisthebestOS
intermsofanythingelse,onlythatithasservedtheirneeds.
Michael
From:julien@incal.inria.fr(JulienMaisonneuve)
Subject:Re:LINUXisobsolete
Date:3Feb9217:10:14GMT
IwouldliketosecondKevinbrowninmostofhisremarks.
I'lladdafewuserpoints:
WhenaststatesthatFSmultithreadingisuseless,itremindsmeofthemany
timesItriedtoletajobruninthebackground(likewhenreadinganarchiveon
afloppy),itisjustunusable,the&shelloperatorcouldevenhavebeenleft
out.
MostinterestingutilitiesarenotevencompilableunderMinixbecauseofthe
ATKcompiler'sincrediblelimits.ThosewerehardlyunderstandableonabasicPC,
butbecomeabsurdona386.EverystupidDOScompilerhasalargemodel(more
expensive,OK).Ihatethe13bitcompress!
ThelackofVirtualMemorysupportpreventspeoplestudyingthisareato
experiment,andpreventsuserstouselargeprograms.Thestrangedesignofthe
MMalsomakesithardtomodify.
Theproblemisthatevendoingexploratoryworkunderminixispainful.
Ifyouwanttogetanyworkdone(orevenfun),evenDOSisbecomingabetter
alternative(withthingslikeDJGPP).
Initsbasicform,itisreallynomorethanOScourseexample,agood
toy,butatoy.Obtainingandapplyingpatchesisapain,andprecludesfurther
upgrades.
Toobadwhennotsomuchismissingtomakeitreallygood.
Thanksfortheworkandy,butLinuxdidn'tdeserveyouranswer.
Forthecommonpeople,itdoesmanythingsbetterthanMinix.
JulienMaisonneuve.
Thisisnotaflame,justmyexperience.
From:richard@aiai.ed.ac.uk(RichardTobin)
Subject:Re:LINUXisobsolete
Date:4Feb9214:46:49GMT
ReplyTo:richard@aiai.UUCP(RichardTobin)
Organization:AIAI,UniversityofEdinburgh,Scotland
http://www.oreilly.com/openbook/opensources/book/appa.html
7/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
Inarticle<12615@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>Amultithreadedfilesystemisonlyaperformancehack.Whenthereisonly
>onejobactive,thenormalcaseonasmallPC,itbuysyounothing
Ifindthesinglethreadedfilesystemaseriouspainwhenusing
Minix.Ioftenwanttodosomethingelsewhilereadingfilesfromthe
(excruciatinglyslow)floppydisk.Iratherliketoplayroguewhile
waitingforlargeCorLispcompilations.Ilooktolookatfilesin
oneeditorbufferwhilecompilinginanother.
(Theproblemwouldbesomewhatlessifthefilesystemstuckto
servingfilesanddidn'tinteractwithterminali/o.)
Ofcourse,inbasicMinixwithnovirtualconsolesandnochanceof
runningemacs,thisisn'tmuchofaproblem.Buttomostpeople
that'safailure,notanadvantage.Itjustisn'tthecasethaton
singleusermachinesthere'snouseformorethanoneactiveprocess;
theideaonlyhasanyplausibilitybecausesomanypeopleareusedto
poormachineswithpooroperatingsystems.
Astoportability,Minixonlywinsbecauseofitslimitedambitions.
IfyouwantedafullfeaturedUnixwithpaging,jobcontrol,awindow
systemandsoon,woulditbequickertostartfrombasicMinixand
addthefeatures,ortostartfromLinuxandfixthe386specific
bits?Idon'tthinkit'sfairtocriticiseLinuxwhenitsaimsareso
differentfromMinix's.Ifyouwantasystemforpedagogicaluse,
Minixistheanswer.Butifwhatyouwantisanenvironmentasmuch
like(say)aSunaspossibleonyourhomecomputer,ithassome
deficiencies.
Richard
From:ast@cs.vu.nl(AndyTanenbaum)
Subject:Re:LINUXisobsolete
Date:5Feb9214:48:48GMT
Organization:Fac.Wiskunde&Informatica,VrijeUniversiteit,Amsterdam
Inarticle<6121@skye.ed.ac.uk>richard@aiai.UUCP(RichardTobin)writes:
>IfyouwantedafullfeaturedUnixwithpaging,jobcontrol,awindow
>systemandsoon,woulditbequickertostartfrombasicMinixand
>addthefeatures,ortostartfromLinuxandfixthe386specific
>bits?
AnotheroptionthatseemstobetotallyforgottenhereisbuyUNIXora
clone.IfyoujustwanttoUSEthesystem,insteadofhackingonits
internals,youdon'tneedsourcecode.Coherentisonly$99,andthere
arevarioustrueUNIXsystemswithmorefeaturesformoremoney.Forthe
truehacker,nothavingsourcecodeisfatal,butforpeoplewhojust
wantaUNIXsystem,therearemanyalternatives(albeitnotfree).
AndyTanenbaum(ast@cs.vul.nl)
From:ajt@doc.ic.ac.uk(TonyTravis)
Subject:Re:LINUXisobsolete
Date:6Feb9202:17:13GMT
Organization:DepartmentofComputing,ImperialCollege,UniversityofLondon,UK.
ast@cs.vu.nl(AndyTanenbaum)writes:
>AnotheroptionthatseemstobetotallyforgottenhereisbuyUNIXora
>clone.IfyoujustwanttoUSEthesystem,insteadofhackingonits
>internals,youdon'tneedsourcecode.Coherentisonly$99,andthere
>arevarioustrueUNIXsystemswithmorefeaturesformoremoney.Forthe
>truehacker,nothavingsourcecodeisfatal,butforpeoplewhojust
>wantaUNIXsystem,therearemanyalternatives(albeitnotfree).
Andy,IhavefollowedthedevelopmentofMinixsincethefirstmessages
werepostedtothisgroupandIamnowrunning1.5.10withBruce
Evans'spatchesforthe386.
I'just'wantaUnixonmyPCandIamnotinterestedinhackingonits
internals,butI*do*wantthesourcecode!
AnimportantprincipleunderlyingthesuccessandpopularityofUnixis
thephilosophyofbuildingontheworkofothers.
Thisphilosophyreliesupontheavailabilityofthesourcecodein
orderthatitcanbeexamined,modifiedandreusedinnewsoftware.
Manyyearsago,IwasinthehappypositionofbeinganAT&TSeventh
EditionUnixsourcelicenceebut,eventhen,Isawyourdecisionto
makethesourceofMinixavailableasliberationfromtheshacklesof
AT&Tcopyright!!
Ithinkyoumaysometimesforgetthatyour'hobby'hashadaprofound
effectontheavailabilityof'personal'Unix(ie.affordableUnix)and
thatthe8086PCIranMinix1.2onactuallycostmeconsiderablymore
thanmypresent386/SXclone.
Clearly,Minix_cannot_beallthingstoallmen,butIseethe
progressto386versionsinmuchthesamewaythatIsee68000orother
http://www.oreilly.com/openbook/opensources/book/appa.html
8/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
linearaddressspacearchitectures:itisagoodthingforpeoplelike
mewhouseMinixandfeelconstrainedbythesegmentedarchitectureof
thePCversionforapplications.
NOTHINGyoucansaywouldconvincemethatIshoulduseCoherent...
Tony
From:richard@aiai.ed.ac.uk(RichardTobin)
Subject:Re:LINUXisobsolete
Date:7Feb9214:58:22GMT
Organization:AIAI,UniversityofEdinburgh,Scotland
Inarticle<12696@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>IfyoujustwanttoUSEthesystem,insteadofhackingonits
>internals,youdon'tneedsourcecode.
Unfortunatelyhackingontheinternalsisjustwhatmanyofuswant
thesystemfor...You'llberidofmostofuswhenBSDdetoxorGNU
comesout,whichshouldhappeninthenextfewmonths(yeah,right).
Richard
From:comm121@unixg.ubc.ca(Louie)
Subject:Re:LINUXisobsolete
Date:30Jan9202:55:22GMT
Organization:UniversityofBritishColumbia,Vancouver,B.C.,Canada
In<12595@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>Butinallhonesty,Iwould
>suggestthatpeoplewhowanta**MODERN**"free"OSlookaroundfora
>microkernelbased,portableOS,likemaybeGNUorsomethinglikethat.
TherearereallynootheralternativesotherthanLinuxforpeoplelike
mewhowanta"free"OS.Consideringthatthemajorityofpeoplewho
wouldusea"free"OSusethe386,portabilityisreallynotallthat
bigofaconcern.IfIhadaSparcIwoulduseSolaris.
Asitstands,IinstalledLinuxwithgcc,emacs18.57,kermitandallofthe
GNUutilitieswithoutanytroubleatall.Noneedtoapplypatches.I
justfollowedtheinstallationinstructions.Ican'tgetanOSlike
this*anywhere*forthepricetodomyComputerSciencehomework.And
itseemslikenetworksupportandthenXWindowswillbeportedtoLinux
wellbeforeMinix.Thisissomethingthatwouldbereallyuseful.Inmy
opinion,portabilityofstandardUnixsoftwareisimportantalso.
Iknowthatthedesignusingamonolithicsystemisnotasgoodasthe
microkernel.Butfortheshorttermfuture(AndIknowIwon't/can't
beupradingfrommy386),Linuxsuitsmeperfectly.
PhilipWu
pwu@unixg.ubc.ca
From:dgraham@bmers30.bnr.ca(DouglasGraham)
Subject:Re:LINUXisobsolete
Date:1Feb9200:26:30GMT
Organization:BellNorthernResearch,Ottawa,Canada
Inarticle<12595@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>WhileIcouldgointoalongstoryhereabouttherelativemeritsofthe
>twodesigns,sufficeittosaythatamongthepeoplewhoactuallydesign
>operatingsystems,thedebateisessentiallyover.Microkernelshavewon.
Canyourecommendany(unbiased)literaturethatpointsoutthestrengths
andweaknessesofthetwoapproaches?I'msurethatthereissomething
tobesaidforthemicrokernelapproach,butIwonderhowclosely
Minixresemblestheothersystemsthatuseit.Sure,Minixuseslots
oftasksandmessages,buttheremustbemoretoamicrokernelarchitecture
thanthat.IsuspectthattheMinixcodeisnotsplitoptimallyintotasks.
>Theonlyrealargumentformonolithicsystemswasperformance,andthere
>isnowenoughevidenceshowingthatmicrokernelsystemscanbejustas
>fastasmonolithicsystems(e.g.,RickRashidhaspublishedpaperscomparing
>Mach3.0tomonolithicsystems)thatitisnowalloverbuttheshoutin`.
MymaincomplaintwithMinixisnotit'sperformance.Itisthatadding
featuresisaroyalpainsomethingthatIpresumeamicrokernel
architecureissupposedtoalleviate.
>MINIXisamicrokernelbasedsystem.
Isthereaconsensusonthis?
>LINUXis
>amonolithicstylesystem.Thisisagiantstepbackintothe1970s.
>Thatisliketakinganexisting,workingCprogramandrewritingitin
http://www.oreilly.com/openbook/opensources/book/appa.html
9/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
>BASIC.Tome,writingamonolithicsystemin1991isatrulypooridea.
Thisisafineassertion,butI'veyettoseeanyrationaleforit.
Linuxisonlyabout12000linesofcodeIthink.Idon'tseehow
splittingthatintotasksandblastingmessagesaroundwouldimproveit.
>Don'tgetmewrong,IamnotunhappywithLINUX.Itwillgetallthepeople
>whowanttoturnMINIXinBSDUNIXoffmyback.Butinallhonesty,Iwould
>suggestthatpeoplewhowanta**MODERN**"free"OSlookaroundfora
>microkernelbased,portableOS,likemaybeGNUorsomethinglikethat.
Well,therearenootherchoicesthatI'mawareofatthemoment.But
whenGNUOScomesout,I'llverylikelyjumpshipagain.Isensethat
you*are*somewhatunhappyaboutLinux(andthatsurprisesmesomewhat).
Iwouldguessthatthereasonsomanypeopleembracedit,isbecauseit
offersmorefeatures.Yourapproachtopeoplerequestingfeaturesin
Minix,hasgenerallybeentotellthemthattheydidn'treallywantthat
featureanyway.IsubmitthattheexodusinthedirectionofLinux
provesyouwrong.
Disclaimer:IhadnothingtodowithLinuxdevelopment.Ijustfind
itaneasiersystemtounderstandthanMinix.
DougGrahamdgraham@bnr.caMyopinionsaremyown.
From:hedrick@klinzhai.rutgers.edu(CharlesHedrick)
Subject:Re:LINUXisobsolete
Date:1Feb9200:27:04GMT
Organization:RutgersUniv.,NewBrunswick,N.J.
Thehistoryofsoftwareshowsthatavailabilitywinsoutover
technicalqualityeverytime.That'sLinux'majoradvantage.It'sa
small386basedsystemthat'sfairlycompatiblewithgenericUnix,and
isfreelyavailable.IdroppedoutoftheMinixcommunityacoupleof
yearsagowhenitbecameclearthat(1)Minixwasnotgoingtotake
advantageofanythingbeyondthe8086anytimeinthenearfuture,and
(2)thelicensingwhileamazinglyfriendlystillmadeithard
forpeoplewhowereinterestedinproducinga386version.Several
peopleapparentlydidniceworkforthe386.Butalltheycould
distributewerediffs.Thismadebringingupa386systemajobthat
isn'tpracticalforanewuser,andinfactIwasn'tsureIwantedto
doit.
Iapologizeifthingshavechangedinthelastcoupleofyears.If
it'snowpossibletogeta386versioninaformthat'sreadytorun,
thecommunityhasdevelopedawaytoshareMinixsource,andbringing
upnormalUnixprogramshasbecomeeasierintheinterim,thenI'm
willingtoreconsiderMinix.Idolikeitsdesign.
It'spossiblethatLinuxwillbeovertakenbyGnuorafreeBSD.
However,iftheGnuOSfollowstheexampleofallotherGnusoftware,
itwillrequireasystemwith128MBofmemoryanda1GBdisktouse.
Therewillstillberoomforasmallsystem.MyidealOSwouldbe4.4
BSD.But4.4'sreleasedatehasahistoryofextremeslippage.With
mostoftheirstaffmovingtoBSDI,it'shardtobelievethatthis
situationisgoingtobeimproved.Formyownpersonaluse,theBSDI
systemwillprobablybegreat.Buteventheirveryattractivepricing
islikelytobetoomuchformostofourstudents,andeventhough
userscangetsourcefromthem,thefactthatsomeofitis
proprietarywillagainmeanthatyoucan'tjustputalteredcodeout
forpublicFTP.Atanyrate,Linuxexists,andtherestofthese
alternativesarevapor.
From:tytso@athena.mit.edu(TheodoreY.Ts'o)
Subject:Re:LINUXisobsolete
Date:31Jan9221:40:23GMT
Organization:MassachusettsInstituteofTechnology
InReplyTo:ast@cs.vu.nl'smessageof29Jan9212:12:50GMT
>From:ast@cs.vu.nl(AndyTanenbaum)
>ftp.cs.vu.nl=192.31.231.42indirminix/simulator.)Ithinkitisa
>grosserrortodesignanOSforanyspecificarchitecture,sincethatis
>notgoingtobearoundallthatlong.
It'snotyourfaultforbelievingthatLinuxistiedtothe80386
architecture,sincemanyLinuxsupporters(includingLinushimself)have
madethethisstatement.However,theamountof80386specificcodeis
probablynotmuchmorethanwhatisinaMiniximplementation,andthere
iscertainlyalotless80386specificcodeinLinuxthanhereis
VaxspecificcodeinBSD4.3.
Granted,theporttootherarchitectureshasn'tbeendoneyet.ButifI
weregoingtobringupaUnixlikesystemonanewarchitecture,I'd
probablystartwithLinuxratherthanMinix,simplybecauseIwantto
havesomecontroloverwhatIcandowiththeresultingsystemwhenI'm
donewithit.Yes,I'dhavetorewritelargeportionsoftheVMand
devicedriverlayersbutI'dhavetodothatwithanyotherOS.
MaybeitwouldbealittlebitharderthanitwouldtoportMinixtothe
newarchitecture;butthiswouldprobablybeonlytrueforthefirst
http://www.oreilly.com/openbook/opensources/book/appa.html
10/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
architecturethatweportedLinuxto.
>WhileIcouldgointoalongstoryhereabouttherelativemeritsofthe
>twodesigns,sufficeittosaythatamongthepeoplewhoactuallydesign
>operatingsystems,thedebateisessentiallyover.Microkernelshavewon.
>Theonlyrealargumentformonolithicsystemswasperformance,andthere
>isnowenoughevidenceshowingthatmicrokernelsystemscanbejustas
>fastasmonolithicsystems(e.g.,RickRashidhaspublishedpaperscomparing
>Mach3.0tomonolithicsystems)thatitisnowalloverbuttheshoutin'.
Thisisnotnecessarilythecase;Ithinkyou'repaintingamuchmore
blackandwhiteviewoftheuniversethannecessarilyexists.Irefer
youtosuchpapersasBrentWelsh's(welch@parc.xerox.com)"The
FilsystemBelongsintheKernel"paper,whereinhearguesthatthe
filesystemisamatureenoughabstractionthatitshouldliveinthe
kernel,notoutsideofitasitwouldinastrictmicrokerneldesign.
Therealsoseveralpeoplewhohavebeenconcernedaboutthespeedof
OSF/1Machwhencomparedwithmonolithicsystems;inparticular,the
nubmerofcontextswitchesrequiredtohandlenetworktraffic,and
networkedfilesystemsinparticular.
Iamawareofthebenefitsofamicrokernelapproach.However,the
factremainsthatLinuxishere,andGNUisn'tandpeoplehavebeen
workingonHurdforalotlongerthanLinushasbeenworkingonLinux.
Minixdoesn'tcountbecauseit'snotfree.:)
Isuspectthatthebalanceofmicrokernelsversusmonolithickernels
dependonwhatyou'redoing.Ifyou'reinterestedindoingresearch,it
isobviouslymucheasiertoripoutandreplacemodulesinamicro
kernel,andsinceonlyresearcherswritepapersaboutoperatingsystems,
ipsofactomicrokernelsmustbetherightapproach.However,Idoknow
alotofpeoplewhoarenotresearchers,butwhoareratherpractical
kernelprogrammers,whohavealotofconcernsoverthecostofcopying
andthecostofcontextswitcheswhichareincurredinamicrokernel.
Bytheway,Idon'tbuyyourargumentsthatyoudon'tneeda
multithreadedfilesystemonasingleusersystem.Onceyoubringupa
windowingsystem,andhaveacompilegoinginonewindow,anewsreader
inanotherwindow,andUUCP/CNewsgoinginthebackground,youwant
goodfilesystemperformance,evenonasingleusersystem.Maybetoa
theoristit'sanunnecessaryoptimizationanda(touseyourwords)
"performancehack",butI'minterestedinaRealoperatingsystem
notaresearchtoy.
===================================
TheodoreTs'obloombeacon!mitathena!tytso
308HighSt.,Medford,MA02155tytso@athena.mit.edu
Everybody'splayingthegame,butnobody'srulesarethesame!
From:joe@jshark.rn.com
Subject:Re:LINUXisobsolete
Date:31Jan9213:21:44GMT
Organization:ablipofentropy
Inarticle<12595@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>
>MINIXwasdesignedtobereasonablyportable,andhasbeenportedfromthe
>Intellinetothe680x0(Atari,Amiga,Macintosh),SPARC,andNS32016.
>LINUXistiedfairlycloselytothe80x86.Notthewaytogo.
Ifyoulookedatthesourceinsteadofbelievingtheauthor,you'drealise
thisisnottrue!
He'sreplaced'fubyte'byaroutinewhichexplicitlyusesasegmentregister
butthatcouldbeeasilychanged.Similarly,apartfromacoupleofplaces
whichassumethe'386MMU,acoupleofmacrostohidetheexactpagesizes
etcwouldmakeportingtrivial.Using'386TSS'smakesthecodesimpler,
buttheVAXandWE32000havesimilarstructures.
Ashe'salreadyadmitted,abitofplanningwouldhavethethesystem
neater,butmerelyputting'386assembleraroundisn'tacrime!
Andwithallduerespect:
theBookdidn'tmakeanissueofportability(apartfromafew
"#ifdefM8088"s)
bythetimeitwasreleased,Minixhadcometodependonseveral
8086"features"thatcauseduproarfromthe68000users.
>AndyTanenbaum(ast@cs.vu.nl)
joe.
From:entropy@wintermute.WPI.EDU(LawrenceC.Foard)
Subject:Re:LINUXisobsolete
Date:5Feb9214:56:30GMT
Organization:WorcesterPolytechnicInstitute
Inarticle<12595@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>Don`tgetmewrong,IamnotunhappywithLINUX.Itwillgetallthepeople
>whowanttoturnMINIXinBSDUNIXoffmyback.Butinallhonesty,Iwould
http://www.oreilly.com/openbook/opensources/book/appa.html
11/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
>suggestthatpeoplewhowanta**MODERN**"free"OSlookaroundfora
>microkernelbased,portableOS,likemaybeGNUorsomethinglikethat.
Ibelieveyouhavesomevalidpoints,althoughIamnotsurethata
microkernelisnecessarilybetter.Itmightmakemoresensetoallowsome
combinationofthetwo.AspartoftheIPCcodeI'mwrittingforLinuxIam
goingtoincludecodethatwillallowdevicedriversandfilesystemstorun
asuserprocesses.Thesewillbesignificantlyslowerthough,andIbelieveit
wouldbeamistaketomoveeverythingoutsidethekernel(TCP/IPwillbe
internal).
ActuallymymainproblemwithOStheoristsisthattheyhavenevertested
thereideas!Noneoftheseideas(withapartialexceptionforMACH)hasever
seenthelightofday.32bithomecomputershavebeenavailableforalmosta
decadeandLinuswasthefirstpersontoeverwriteaworkingOSforthem
thatcanbeusedwithoutpayingAT&T$100,000.Apieceofsoftwareinhandis
worthtenpiecesofvaporware,OStheoristsarequicktojumpalloveranOS
buttheyareunwillingtoeverprovideanalternative.
ThegeneralconsensusthatMicrokernelsisthewaytogomeansnothingwhen
arealapplicationhasneverevenrunonone.
ThereleaseofLinuxisallowingmetotrysomeideasI'vebeenwantingto
expermentwithforyears,butIhaveneverhadtheopportunitytoworkwith
sourcecodeforafunctioningOS.
From:ast@cs.vu.nl(AndyTanenbaum)
Subject:Re:LINUXisobsolete
Date:5Feb9223:33:23GMT
Organization:Fac.Wiskunde&Informatica,VrijeUniversiteit,Amsterdam
Inarticle<1992Feb5.145630.759@wpi.WPI.EDU>entropy@wintermute.WPI.EDU(Lawrence
C.Foard)writes:
>ActuallymymainproblemwithOStheoristsisthattheyhavenevertested
>thereideas!
I'mmortallyinsulted.IAMNOTATHEORIST.Askanybodywhowasatour
departmentmeetingyesterday(injoke).
Actually,theseideashavebeenverywelltestedinpractice.OSFisbetting
itswholebusinessonamicrokernel(Mach3.0).USLisbettingitsbusiness
onanotherone(Chorus).Bothoftheserunlotsofsoftware,andbothhave
beenextensivelycomparedtomonolithicsystems.Amoebahasbeenfully
implementedandtestedforanumberofapplications.QNXisamicrokernel
basedsystem,andsomeonejusttoldmetheinstalledbaseis200,000systems.
Microkernelsarenotapipedream.Theyrepresentproventechnology.
TheMachguyswroteapapercalled"UNIXasanapplicationprogram."
ItwasbyGolubetal.,intheSummer1990USENIXconference.TheChorus
peoplealsohaveatechnicalreportonmicrokernelperformance,andI
coauthoredanotherpaperonthesubject,whichImentionedyesterday
(Dec.1991ComputingSystems).Checkthemout.
AndyTanenbaum(ast@cs.vu.nl)
From:peter@ferranti.com(peterdasilva)
Subject:Re:LINUXisobsolete
Organization:XenixSupport,FICC
Date:Thu,6Feb199216:02:47GMT
Inarticle<12747@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>QNXisamicrokernel
>basedsystem,andsomeonejusttoldmetheinstalledbaseis200,000systems.
Ohyes,whileI'monthesubject...thereareover3millionAmigasoutthere,
whichmeansthattherearemoreofthemthananyUNIXvendorhasshipped,and
probablymorethanallUNIXsystemscombined.
From:peter@ferranti.com(peterdasilva)
Subject:Re:LINUXisobsolete
Organization:XenixSupport,FICC
Date:Thu,6Feb199216:00:22GMT
Inarticle<1992Feb5.145630.759@wpi.WPI.EDU>entropy@wintermute.WPI.EDU(Lawrence
C.Foard)writes:
>ActuallymymainproblemwithOStheoristsisthattheyhavenevertested
>thereideas!
Ibegtodiffer...therearemanymicrokerneloperatingsystemsoutthere
foreverythingfroman8088(QNX)uptolargeresearchsystems.
>Noneoftheseideas(withapartialexceptionforMACH)hasever
>seenthelightofday.32bithomecomputershavebeenavailableforalmosta
>decadeandLinuswasthefirstpersontoeverwriteaworkingOSforthem
>thatcanbeusedwithoutpayingAT&T$100,000.
ImusthavebeenimaginingAmigaOS,then.I'vebeenusingafigmentofmy
imaginationforthepast6years.
http://www.oreilly.com/openbook/opensources/book/appa.html
12/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
AmigaOSisamicrokernelmessagepassingdesign,withbetterresponsetime
andperformancethananyotherreadilyavailablePCoperatingsystem:including
MINIX,OS/2,Windows,MacOS,Linux,UNIX,and*certainly*MSDOS.
Themicrokerneldesignhasproveninvaluable.Thingslikenewfilesystems
thatarenormallyavailableonlyfromthevendorarehobbyistproductson
theAmiga.Devicedriversaresimplysharedlibrariesandtaskswithspecific
entrypointsandmessageports.Soarefilesystems,thewindowsystem,and
soon.It'saWONDERFULdesign,andvalidateseverythingthatpeoplehave
beensayingaboutmicrokernels.Yes,ittakesmoreworktogetthemoffthe
groundthanacoroutinebasedmacrokernellikeUNIX,buttheversatility
paysyoubackmanytimesover.
IreallywishAndywoulddoanewMINIXbasedonwhathasbeenlearnedsince
thefirstrelease.ThefactoringofresponsibilitiesinMINIXisfairlypoor,
butthebasicconceptisgood.
>ThegeneralconsensusthatMicrokernelsisthewaytogomeansnothingwhen
>arealapplicationhasneverevenrunonone.
I'mdreamingagain.IsurethroughtDeluxePaint,Sculpt3d,PhotonPaint,
ManxC,ManxSDB,PerfectSound,Videoscape3d,andtheotherprogramsI
boughtformyAmigawere"real".I'llhavetosendthedamnthingsbacknow,
Iguess.
TheavailabilityofLinuxisgreat.I'mdelighteditexists.I'msurethat
themacrokerneldesignisonereasonithasbeenimplementedsofast,andthis
isavalidreasontousemacrokernels.BUT...thisdoesn'tmeanthat
microkernelsareinherentlyslow,orsimplyresearchtoys.
From:dsmythe@netcom.COM(DaveSmythe)
Subject:Re:LINUXisobsolete
Date:10Feb9207:08:22GMT
Organization:NetcomOnlineCommunicationServices(4082419760guest)
Inarticle<1992Feb5.145630.759@wpi.WPI.EDU>entropy@wintermute.WPI.EDU(Lawrence
C.Foard)writes:
>ActuallymymainproblemwithOStheoristsisthattheyhavenevertested
>thereideas!Noneoftheseideas(withapartialexceptionforMACH)hasever
>seenthelightofday.
DavidCheriton(Prof.atStanford,andauthoroftheVsystem)saidsomething
similartothisinaclassindistributedsystems.Paraphrased:
"Therearetwokindsofresearchers:thosethathaveimplemented
somethingandthosethathavenot.Thelatterwilltellyouthat
thereare142waysofdoingthingsandthatthereisn'tconsensus
onwhichisbest.Theformerwillsimplytellyouthat141of
themdon'twork."
HereallyripsontheOSIphilesaswell,forasimilarreason.TheInternet
protocolsareadaptedonlyafterhavingbeeninuseforaperiodoftime,
preventingthingsfromgettingstandardizedthatwillneverbeimplementable
inareasonablefashion.OSIadherents,ontheotherhand,seemintenton
standardizingeverythingpossible,including"escapes"fromthestandard,
beforeareasonablereferenceimplementationexists.Consequently,yousee
obsoleteideasimmortalized,suchassubbyteleveldatafieldpacking,
whichmakesgoodperformancedifficultwhenyourcomputerisdrinkingfrom
a10+Gbsfirehose:).
Justmy$.02
From:torvalds@klaava.Helsinki.FI(LinusBenedictTorvalds)
Subject:Apologies(wasRe:LINUXisobsolete)
Date:30Jan9215:38:16GMT
Organization:UniversityofHelsinki
Inarticle<1992Jan29.231426.20469@klaava.Helsinki.FI>Iwrote:
>Well,withasubjectlikethis,I'mafraidI'llhavetoreply.
AndreplyIdid,withcompleteabandon,andnothoughtforgoodtaste
andnetiquette.Apologiestoast,andthankstoJohnNallforafriendy
"that'snothowit'sdone"letter.Ioverreacted,andamnowcomposing
a(muchlessacerbic)personallettertoast.Hopenobodywasturned
awayfromlinuxduetoitbeing(a)possiblyobsolete(Istillthink
that'snotthecase,althoughsomeofthecriticismsarevalid)and(b)
writtenbyahothead:)
Linus"myfirst,andhopefullylastflamefest"Torvalds
From:pmacdona@sanjuan(PeterMacDonald)
Subject:re:Linuxisobsolete
Date:1Feb9202:10:06GMT
Organization:UniversityofVictoria,Victoria,BC,CANADA
SinceIthinkIpostedoneoftheearliestmessagesinallthisdiscussion
http://www.oreilly.com/openbook/opensources/book/appa.html
13/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
ofMinixvsLinux,Ifeelcompelledtocommentonmyreasonsfor
switchingfromMinixtoLinux.Inorderofimportancetheyare:
1)Linuxisfree
2)Linuxisevolvingatasatisfactoryclip(becausenewfeatures
areacceptedintothedistributionbyLinus).
Thefirstrequiressomeexplanation,becauseifIhavealreadypurchased
Minix,whatposssibleconcerncouldpricehaveforme?Simple.
IftheOSisfree,manymorepeoplewilluse/support/enhanceit.
ThisisalsothesamereasoningIusedwhenIboughtmy386instead
ofasparc(whichIcouldhavegotforjust30%more).Since
PCsarecheapandgenerallyavailable,morepeoplewillbuy/use
themandthusgood,cheap/freesoftwarewillbeabundant.
ThesecondshouldbeprettyobvioustoanyonewhohasbeenusingMinix
forforanyperiodoftime.ASTgenerallydoesnotacceptenhancements
toMinix.Thisisnotmeantasachallenge,butmerelyastatementof
fact.ASThasgoodandlegitimatereasonsforthis,andIdonotdispute
them.ButMinixhassomelimitationswhichIjustcouldnolonger
livewith,andduetothispolicy,theprospectofseeingthemresolved
inreasonabletimewasunsatisfactory.Theselimitationsinclude:
no386support
novirtualconsoles
nosoftlinks
noselectcall
noptys
nodemandpaging/swapping/sharedtext/sharedlibs...(efficientmm)
chmem(inflexiblemm)
noXWindows(advocatedforthesamereasonsasLinuxandthe386).
noTCP/IP
noGNU/SysVintegration(portability)
Someofthesecouldbefixedbypatches(andifyouhavedonethis
yourself,Idon'thavetotellyouhowsatisfactorythatis),butat
leastthelast5itemswere/arebeyondanyreasonableexpectation.
Finally,mycomment(crack?)aboutMinix'ssegmentedkernel,or
microkernelarchitecturewasmoreanexpressionofmyfrustration/
bewildermentatattemptingtousetheMinixPTYpatchesasaguide
ofhowtodoitunderLinux.Thatparticularinstancewasonewhere
messagepassinggreatlycomplicatedtheimplementationofafeature.
IdohaveanopinionaboutMonlithicvsMessagePassing,butwon't
expressitnow,anddidnotmeantoexpresssitthen.Mygoalsare
totallyshortterm(maximumfunctionalityintheminimumamountof
time/cost/hassle),andsomyviewsonthisareirrelevant,andshould
notbemisconstrued.Ifyouarenonplussedbythelackoftheabove
features,thenyoushouldconsiderMinix,aslongasyoudon'tmind
payingofcourse:)
From:olaf@oski.toppoint.de(OlafSchlueter)
Subject:Re:Linuxisobsolete
Date:7Feb9211:41:44GMT
Organization:ToppointMailboxe.V.
JustafewcommentstothediscussionofLinuxvsMinix,whichevolved
partlytoadiscussionofmonolithicvsmicrokernel.
Ithinktherewillbenoaggreementbetweenthetwopartiesadvocating
eitherconcept,iftheyforget,thatLinuxandMinixhavebeendesigned
fordifferentapplications.Ifyouwantacheap,powerfuland
enhancableUnixsystemrunningonasinglemachine,withthepossibility
toadaptstandardUnixsoftwarewithoutpain,thenLinuxisforyou.If
youareinterestedinmodernoperatingsystemconcepts,andwantto
learnhowamicrokernelbasedsystemworks,thenMinixisthebetter
choice.
Itisnotanargumentagainstmicrokernelsystem,thatforthetime
beingmonolithicimplemenationsofUnixonPCshaveabetter
performance.Thismeansonly,thatUnixismaybebetterimplementedas
amonolithicOS,atleastaslongasitrunsonasinglemachine.From
theuserspointofview,theinternaldesignoftheOSdoesn'tmatterat
all.Untilitcomestonetworks.Onthemonolithicapproach,afile
serverwillbecomeauserprocessbasedonsomehardwarefacilitylike
ethernet.Programswhichwanttousethisfacilitywillhavetouse
speciallibrarieswhichofferthecallsforcommunicationwiththis
server.Inamicrokernelsystemitispossibletoincorporatethe
serverintotheOSwithouttheneedfornew"system"calls.Fromthe
userspointofviewthishastheadvantage,thatnothingchanges,he
justgetsbetterperformance(intermsofmorediskspaceforexample).
Fromtheimplementorspointofview,themicrokernelsystemisfaster
adaptabletochangesinhardwaredesign.
Ithasbeencritized,thatASTrejectsanyimprovementstoMinix.Ashe
isinterestedintheeducationalvalueofMinix,Iunderstandhis
argument,thathewantstokeepthecodesimple,anddon'twantto
overloaditwithfeatures.Asaneducationaltool,Minixiswrittenas
amicrokernelsystem,althoughitisrunningonhardwareplatforms,who
willprobablybetterperformwithamonolithicOS.Buttheareaof
networkapplicationsisgrowingandmodernOSlikeAmoebaorPlan9
http://www.oreilly.com/openbook/opensources/book/appa.html
14/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
cannotbewrittenasmonolithicsystems.SoMinixhasbeenwrittenwith
theintentiontogivestudentsapracticalexampleofamicrokernelOS,
toletthemplaywithtasksandmessages.Itwasnottheideatogivea
lotofpeopleacheap,powerfulOSforatenthofthepriceofSYSVor
BSDimplementations.
Resumee:LinuxisnotbetterthanMinix,ortheotherwayround.They
aredifferentforgoodreasons.
From:meggin@epas.utoronto.ca(DavidMegginson)
Subject:Mach/Minix/Linux/Gnuetc.
Date:1Feb9217:11:03GMT
Organization:UniversityofTorontoEPAS
Well,thishasbeenafundiscussion.Iamabsolutelyconvincedby
Prof.Tanenbaumthatamicrokernel_is_thewaytogo,butthemore
IlookattheMinixsource,thelessIbelievethatitisa
microkernel.IwouldprobablynotbotherportingLinuxtothe
M68000,butIwantmoreservicesthanMinixcanoffer.
Whataboutamicrokernelwhichismessage/syscallcompatiblewith
MACH?Itdoesn'tactuallyhavetodoeverythingthatMACHdoes,like
virtualmemorypagingitjusthasto_look_likeMACHfromthe
outside,tofoolprogramslikethefutureGnuUnixemulator,BSD,etc.
ThiswouldextendtheusefullivesofourM68000or80286based
machinesforalittlelonger.Inthemeantime,Iwillprobablystay
withMinixformySTratherthanswitchingbacktoMiNTafterall,
MinixatleastlookslikeUnix,whileMiNTlookslikeTOStryingto
looklikeUnix(ithasto,tobeTOScompatible).
David
From:peter@ferranti.com(peterdasilva)
Newsgroups:comp.os.minix
Subject:Whatgooddoesthiswardo?(Re:LINUXisobsolete)
Date:3Feb9216:37:24GMT
Organization:XenixSupport,FICC
Willyouquitflamingeachother?
Imean,linuxisdesignedtoprovideareasonablyhighperformanceenvironment
onahardwareplatformcrippledbyyearsofbackwardscompatiblekludges.Minix
isdesignedasateachingtool.Neitheristhatgoodatdoingtheother'sjob,
andwhyshouldthey?ThefactthatMinixrunsoutofsteamquickly(andit
does)isn'taprobleminitschosenmileau.It'ssurebetterthantheTOY
operatingsystem.ThefactthatLinuxisn'ttransportablebeyondthe386/AT
platformisn'taproblemwhentherearemillionsofthemoutthere(andquite
cheap:youcangeta386/SXforwellunder$1000).
Amonolithickerneliseasyenoughtobuildthatit'sworthdoingitifitgets
asystemoutthedoorearly.Thinkofitasaperformancehackforprogrammer
time.TheAPIisportable.Youcanreplacethekernelwithamicrokernel
design(andMINIXisn'tthebeallandendallofmicrokerneldesignseither:
evenforlowendPCs...lookatAmigaOS)withoutdisturbingtheapplications.
That'sthewholepointofaportableAPIinthefirstplace.
Microkernelsaredefinitelyabetterdesignformanytasks.Itakesmore
worktomakethemefficient,soasimplerdesignthatdoesn'ttakeadvantage
ofthemicrokernelinanyrealwayisworthdoingforpedagogicalreasons.
Thinkofitasaperformancehackforstudenttime.Thedesignisstillgood
andwhenyoucangetanAPItothemicrokernelinterfaceyoucangetVERY
impressiveperformance(thousandsofcontextswitchespersecondonan8
MHz68000).
From:ast@cs.vu.nl(AndyTanenbaum)
Subject:Unhappycampers
Date:3Feb9222:46:40GMT
Organization:Fac.Wiskunde&Informatica,VrijeUniversiteit,Amsterdam
I'vebeengettingabitofmaillatelyfromunhappycampers.(Actually10
messagesfromthe43,000readersmayseemlikealot,butitisnotreally.)
Thereseemtobethreestickingpoints:
1.Monolithickernelsarejustasgoodasmicrokernels
2.Portabilityisn'tsoimportant
3.Softwareoughttobefree
Ifpeoplewanttohaveaseriousdiscussionofmicrokernelsvs.monolithic
kernels,fine.Wecandothatincomp.os.research.Butpleasedon'tsoundoff
ifyouhavenoideaofwhatyouaretalkingabout.Ihavehelpeddesign
andimplement3operatingsystems,onemonolithicandtwomicro,andhave
studiedmanyothersindetail.Manyoftheargumentsofferedarenonstarters
(e.g.,microkernelsarenogoodbecauseyoucan'tdopaginginuserspace
exceptthatMachDOESdopaginginuserspace).
Ifyoudon'tknowmuchaboutmicrokernelsvs.monolithickernels,thereis
someusefulinformationinapaperIcoauthoredwithFredDouglis,Frans
KaashoekandJohnOusterhoutintheDec.1991issueofCOMPUTINGSYSTEMS,the
http://www.oreilly.com/openbook/opensources/book/appa.html
15/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
USENIXjournal).Ifyoudon'thavethatjournal,youcanFTPthepaperfrom
ftp.cs.vu.nl(192.31.231.42)indirectoryamoeba/papersascomp_sys.tex.Z
(compressedTeXsource)orcomp_sys.ps.Z(compressedPostScript).Thepaper
givesactualperformancemeasurementsandsupportsRickRashid'sconclusionthat
microkernelbasedsystemsarejustasefficientasmonolithickernels.
Astoportability,thereishardlyanyseriousdiscussionpossibleanymore.
UNIXhasbeenportedtoeverythingfromPCstoCrays.Writingaportable
OSisnotmuchharderthananonportableone,andallsystemsshouldbe
writtenwithportabilityinmindthesedays.SurelyLinus'OSprofessor
pointedthisout.MakingOScodeportableisnotsomethingIinventedin1987.
Whilemostpeoplecantalkrationallyaboutkerneldesignandportability,
theissueoffreenessis100%emotional.Youwouldn'tbelievehowmuch
[expletivedeleted]IhavegottenlatelyaboutMINIXnotbeingfree.MINIX
costs$169,butthelicenseallowsmakingtwobackupcopies,sotheeffective
pricecanbeunder$60.Furthermore,professorsmaymakeUNLIMITEDcopies
fortheirstudents.Coherentis$99.FSFcharges>$100forthetapeits"free"
softwarecomesonifyoudon'thaveInternetaccess,andIhaveneverheard
anyonecomplain.4.4BSDis$800.Idon'treallybelievemoneyistheissue.
Besides,probablymostofthepeoplereadingthisgroupalreadyhaveit.
ApointwhichIdon'tthinkeveryoneappreciatesisthatmakingsomething
availablebyFTPisnotnecessarilythewaytoprovidethewidestdistribution.
TheInternetisstillahighlyelitegroup.MostcomputerusersareNOTonit.
ItismyunderstandingfromPHthatthecountrywhereMINIXismostwidelyused
isGermany,nottheU.S.,mostlybecauseoneofthe(commercial)German
computermagazineshasbeenactivelypushingit.MINIXisalsowidelyusedin
EasternEurope,Japan,Israel,SouthAmerica,etc.Mostofthesepeoplewould
neverhavegottenitiftherehadn'tbeenacompanysellingit.
Gettingbacktowhat"free"means,whataboutfreesourcecode?Coherent
isbinaryonly,butMINIXhassourcecode,justasLINUXdoes.Youcanchange
itanywayyouwant,andpostthechangeshere.Peoplehavebeendoingthat
for5yearswithoutproblems.Ihavebeengivingfreeupdatesforyears,too.
Ithinktherealissueissomethingelse.I'vebeenrepeatedlyofferedvirtual
memory,paging,symboliclinks,windowsystems,andallmanneroffeatures.I
haveusuallydeclinedbecauseIamstilltryingtokeepthesystemsimple
enoughforstudentstounderstand.Youcanputallthisstuffinyourversion,
butIwon'tputitinmine.Ithinkitisthispointwhichirksthepeoplewho
say"MINIXisnotfree,"notthe$60.
AninterestingquestioniswhetherLinusiswillingtoletLINUXbecome"free"
ofhiscontrol.Maypeoplemodifyit(ruinit?)andsellit?Rememberthe
hundredsofmessageswithsubject"Re:Yoursoftwaresoldformoney"whenit
wasdiscoveredtheMINIXCentreinEnglandwassellingdisketteswithnews
postings,moreorlessatcost?
SupposeFredvanKempenreturnsfromthedeadandwantstotakeover,creating
Fred'sLINUXandLinus'LINUX,bothusefulbutdifferent.Isthatok?The
testcomeswhenasizablegroupofpeoplewanttoevolveLINUXinawayLinus
doesnotwant.Untilthatactuallyhappensthepointismoot,however.
IfyoulikeLinus'philosophyratherthanmine,byallmeans,followhim,but
pleasedon'tclaimthatyou'redoingthisbecauseLINUXis"free."Just
saythatyouwantasystemwithlotsofbellsandwhistles.Fine.Yourchoice.
Ihavenoargumentwiththat.Justtellthetruth.
Asanaside,forthosefolkswhodon'treadnewsheaders,LinusisinFinland
andIaminTheNetherlands.Arewereachingasituationwhereanother
criticalindustry,freesoftware,thathadbeentotallydominatedbytheU.S.
isbeingtakenoverbytheforeigncompetition?Willwesoonsee
PresidentBushcomingtoEuropewithRichardStallmanandRickRashid
intow,demandingthatEuropeimportmoreAmericanfreesoftware?
AndyTanenbaum(ast@cs.vu.nl)
From:ast@cs.vu.nl(AndyTanenbaum)
Subject:Re:Unhappycampers
Date:5Feb9223:23:26GMT
Organization:Fac.Wiskunde&Informatica,VrijeUniversiteit,Amsterdam
Inarticle<205@fishpond.uucp>fnf@fishpond.uucp(FredFish)writes:
>IfPHwasnotgrantedamonopolyondistribution,itwouldhavebeenpossible
>foralloftheinterestedminixhackerstoorganizeandsetupagroupthat
>wasdedicatedtoproducingenhancedminix.Thisaimofthisgroupcouldhave
>beentoproduceasingle,supportedversionofminixwithallofthecommonly
>requestedenhancements.Thiswouldhaveallowedminixtoevolveinmuchthe
>samewaythatgcchasevolvedoverthelastfewyears.
ThisISpossible.Ifagroupofpeoplewantstodothis,thatisfine.
Ithinkcoordinating1000primadonnaslivingallovertheworldwillbe
aseasyasherdingcats,butthereisnolegalproblem.Whenanewrelease
isready,justmakeadifflistingagainst1.5andpostitormakeitFTPable.
Whilethiswillrequiresomeworkonthepartoftheuserstoinstallit,
itisn'tthatmuchwork.Besides,Ihaveshellscriptstomakethediffs
andinstallthem.ThisiswhatFredvanKempenwasdoing.Whathedidwrong
wasinsistontherighttopublishthenewversion,ratherthandiffsagainst
thePHbaseline.ThatcutsPHoutoftheloop,which,notsurprisingly,they
weren'twildabout.Ifpeoplestillwanttodothis,goahead.
Ofcourse,Iamnotnecessarilygoingtoputanyofthesechangesinmyversion,
http://www.oreilly.com/openbook/opensources/book/appa.html
16/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
sothereissomeworkkeepingtheofficialandenhancedonesinsync,butI
amwillingtocooperatetominimizework.Ididthisforalongtimewith
BruceEvansandFransMeulenbroeks.
IfLinuswantstokeepcontroloftheofficialversion,andagroupofeager
beaverswanttogooffinadifferentdirection,thesameproblemarises.
Idon'tthinkthecopyrightissueisreallytheproblem.Theproblemis
coordinatingthings.ProjectslikeGNU,MINIX,orLINUXonlyholdtogether
ifonepersonisincharge.Duringthe1970s,whenstructuredprogramming
wasintroduced,HarlanMillspointedoutthattheprogrammingteamshould
beorganizedlikeasurgicalteamonesurgeonandhisorherassistants,
notlikeahogbutcheringteamgiveeverybodyanaxeandletthemchopaway.
Anyonewhosaysyoucanhavealotofwidelydispersedpeoplehackawayon
acomplicatedpieceofcodeandavoidtotalanarchyhasnevermanageda
softwareproject.
>Whereisthesizeablegroupofpeoplethatwanttoevolvegccinawaythat
>rms/FSFdoesnotapproveof?
Acompilerisnotsomethingpeoplehavemuchemotionalattachmentto.If
thelanguagetobecompiledisagiven(e.g.,anANSIstandard),thereisn't
muchroomforpeopletoinventnewfeatures.Anoperatingsystemhasunlimited
opportunityforpeopletoimplementtheirownfavoritefeatures.
AndyTanenbaum(ast@cs.vu.nl)
From:torvalds@klaava.Helsinki.FI(LinusBenedictTorvalds)
Subject:Re:Unhappycampers
Date:6Feb9210:33:31GMT
Organization:UniversityofHelsinki
Inarticle<12746@star.cs.vu.nl>ast@cs.vu.nl(AndyTanenbaum)writes:
>
>IfLinuswantstokeepcontroloftheofficialversion,andagroupofeager
>beaverswanttogooffinadifferentdirection,thesameproblemarises.
ThisisthesecondtimeI'veseenthis"accusation"fromast,whofeels
prettygoodaboutcommentingonakernelheprobablyhaven'tevenseen.
Oratleasthehasn'taskedme,orevenreadalt.os.linuxaboutthis.
Justsothatnobodytakeshisguessforthefullthruth,here'smy
standingon"keepingcontrol",in2words(three?):
Iwon't.
TheonlycontrolI'veeffectivelybeenkeepingonlinuxisthatIknow
itbetterthananybodyelse,andI'vemademychangesavailableto
ftpsitesetc.Thosehavebecomeeffectivelyofficialreleases,andI
don'texpectthistochangeforsometime:notbecauseIfeelIhave
somemoralrighttoit,butbecauseIhaven'theardtoomanycomplaints,
anditwillbeacoupleofmonthsbeforeIexpecttofindpeoplewho
havethesame"feel"forwhathappensinthekernel.(Well,maybe
peoplearegettingthere:tytsocertainlymadesomeheavychangeseven
to0.10,andothershavehackeditaswell)
InfactIhavesentoutfeelersaboutsome"linuxkernel"mailinglist
whichwouldmakethedecisionsaboutreleases,asIexpectIcannot
fullysupportallthefeaturesthatwill/have/tobeadded:SCSIetc,
thatIdon'thavethehardwarefor.Theresponsehasbeennonexistant:
peopledon'tseemtobethateagertochangeyet.(well,oneperson
feltIshouldaskaroundfordonationssothatIcouldsupportitand
ifanybodyhasinterestinghardwarelyingaround,I'dbehappytoaccept
it:)
Theonlythingthecopyrightforbids(andIfeelthisiseminently
reasonable)isthatotherpeoplestartmakingmoneyoffit,anddon't
makesourceavailableetc...Thismaynotbeaquestionoflogic,but
I'dfeelverybadifsomeonecouldjustsellmyworkformoney,whenI
madeitavailableexpresslysothatpeoplecouldplayaroundwitha
personalproject.Ithinkmostpeopleseemypoint.
Thataside,ifFredvanKempenwantedtomakeasuperlinux,he'squite
wellcome.Hewon'tbeabletomakemuchmoneyonit(distributionfee
only),andIdon'tthinkit'sthatgoodanideatosplitlinuxup,butI
wouldn'twanttostophimevenifthecopyrightletme.
>Idon'tthinkthecopyrightissueisreallytheproblem.Theproblemis
>coordinatingthings.ProjectslikeGNU,MINIX,orLINUXonlyholdtogether
>ifonepersonisincharge.
Yes,coordinationisabigproblem,andIdon'tthinklinuxwillmove
awayfrommeas"headsurgeon"forsometime,partlybecausemostpeople
understandabouttheseproblems.Butcopyright/is/anissue:ifpeople
feelIdoabadjob,theycandoitthemselves.Likewisewithgcc.The
minixcopyright,however,meansthatifsomeonefeelshecouldmakea
betterminix,heeitherhastomakepatches(whicharen'tthatgreat
whateveryousayaboutthem)orstartofffromscratch(andbeattacked
becauseyouhaveotherideals).
Patchesaren'tmuchfuntodistribute:Ihaven'tmadecdiffsfora
singleversionoflinuxyet(Iexpectthistochange:soonthepatches
willbesomuchsmallerthanthekernelthatmakingbothpatchesanda
completeversionavailableisagoodideanotethatI'dstillmakethe
http://www.oreilly.com/openbook/opensources/book/appa.html
17/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
wholeversionavailabletoo).Patchesuponpatchesaresimply
impractical,especiallyforpeoplethatmaydochangesthemselves.
>>Whereisthesizeablegroupofpeoplethatwanttoevolvegccinawaythat
>>rms/FSFdoesnotapproveof?
>Acompilerisnotsomethingpeoplehavemuchemotionalattachmentto.If
>thelanguagetobecompiledisagiven(e.g.,anANSIstandard),thereisn't
>muchroomforpeopletoinventnewfeatures.Anoperatingsystemhasunlimited
>opportunityforpeopletoimplementtheirownfavoritefeatures.
Well,there'sGNUemacs...Don'ttelluspeoplehaven'tgotemotional
attachmenttoeditors:)
Linus
From:dmiller@acg.uucp(DavidMiller)
Subject:LinuxisObsoleteandfollowuppostings
Date:3Feb9201:03:46GMT
Organization:AppliedComputerGroup
Asanobserverinterestedinoperatingsystemdesign,Icouldn'tresistthis
thread.PleaserealizethatIamnotreallyexperiencedwithminux
orlinux:Ihavebeenintounixformanyyears.First,afewobservations:
MinixwaswrittentobeaneducationaltoolforASTs'classes,notacommercial
operatingsystem.Itwasneveradesignparametertohaveitrunfreely
availablesourcecodeforunixsystems.Ithinkitwasalsoastatementof
howoperatingsystemsshouldbedesigned,withamicrokernelandseperate
processescoveringasmuchoftherequiredfunctionalityaspossible.
LinuxwaswrittenmostlyasalearningexerciseonLinusparthowto
programthe386family.Designingtheultimateoperatingsystemwasnot
anobjective.Providingausable,freeplatformthatwouldrunallsorts
ofwidelyavailablefreesoftwarewasaconsideration,andonethatappears
tohavebeenwellmet.
Criticismfromanyonethateitherofthesesystemsisn'twhat*they*would
likeittobeismisplaced.Afterall,anybodythathasacomputerthatwill
runeithersystemisfreetodowhatLinusandAndrewdid:writeyourown!
I,forone,applaudLinusforhisconsiderableeffortindevelopingLinux
andhisdecisiontomakeitfreetoeverybody.IapplaudASTforhis
efforttomakeminixaffordableIhaverealtroublerelatingtocomplaints
thatminixisn'tfree.Ifyoucanaffordthetimetoexploreminix,anda
basiccomputersystem,$150isnotmuchmoreandyoudogetabooktogo
withit.
Next,afewquestionsfortheprofessor:
Isminixsupposedtobea"realoperatingsystem"oraneducationaltool?
Asaneducationaltoolitisanexcellentwork.Asarealoperatingsystem
itpresentssometerriblyroughedges(whynomalloc()?,justforstarters)
MyfeelingfromreadingTheBookandlisteningtopostingshereisthatyou
wantedatooltoteachyourclasses,andalotofotherswantedtoplaywith
anaffordableoperatingsystem.Theseothershavebeentryingtobolton
enoughfeaturestomakeita"realoperatingsystem",withlessthan
outstandingsuccess.
Whysplitfundementalosfunctions,suchasmemorymanagement,intouser
processes?Asallgood*nixgurusknow,themeanstosuccessisto
divideandconquer,withthegoalbeingto*simplify*theprobleminto
managable,welldefinedcomponents.Ifsplittingbasicpartsofthe
operatingsystemintouserspaceprocessescomplicatesthefunctionby
introducingadditionalmechanisms(messagepassing,complicatedsignals),
havewemettheobjectiveofsimplifyingthedesignandimplementation?
Iagreethat*nixhassufferedabadcaseoffeatureitisespecially
sysVr4.Perhapsthefeaturesthatpeoplewantforeitherfunctionality
orcompatibilitycouldbeofferedbyruntimeloadablemodules/libraries
thatofferthesefeatures.Themicrokernelwouldstillbeabaselevel
resourcemanagerthatalsoroutesfunctionrequeststotheappropriate
module/library.Themodulescouldbethreadsoruserprocesses.(Ithink
oshackerspleasecorrectme:))
Justmy$.04worthpleasefeelfreetopostoremailresponses.
Ihavenoformalprogressivetrainingincomputerscience,soIamreally
askingthesequestionsinignorance.Isuspectalotofothersonthe
nethavesimilarquestionsintheirownminds,butI'vebeenwrongbefore.
David
From:michael@gandalf.informatik.rwthaachen.de(MichaelHaardt)
Subject:1.6.17summaryandwhyIthinkASTisright.
Date:6Feb9220:07:25GMT
ReplyTo:u31b3hs@messua.informatik.rwthaachen.de(MichaelHaardt)
Organization:Gandalfa38620machine
IwillfirstgiveasummaryofwhatyoucanexpectfromMINIXin*near*
future,andthenexplainwhyIthinkASTisright.
http://www.oreilly.com/openbook/opensources/book/appa.html
18/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
Sometimeago,IaskedfordetailsaboutthenextMINIXrelease(1.6.17).
Igotsomeresponse,butonlyfrompeoplerunning1.6.16.Thefollowing
informationsarenotofficialandmaybewrong,buttheyareallIknow
atthemoment.Correctmeifsomethingiswrong:
The1.6.17patcheswillberelativeto1.5asshippedbyPH.
Theheaderfilesareclean.
Thetwotypesoffilesystemscanbeusedtogether.
ThesignalhandlingisrewrittenforPOSIX.Theoldbugisremoved.
TheANSIcompiler(availablefromTransmediar,Iguess)comeswith
compilerbinariesandnewlibraries.
Theredon'tseemtobesupportfortheAmoebanetworkprotocol.
times(2)returnsacorrectvalue.termios(2)isimplemented,butit's
moreahack.Idon'tknowif"implemented"meansinthekernel,orthe
currentemulation.
Thereisnodocumentationaboutthenewfilesystem.Thereisanewfsck
andanewmkfs,don'tknowaboutde.
WiththeANSIcompiler,thereisbetterfloatingpointsupport.
Theschedulerisimproved,butnotasgoodaswrittenbyKaiUweBloem.
IaskedthesethingstogetfactsforthedecisionifIshouldupgradeto
MINIX1.6.17ortoLinuxaftertheexamensareover.Well,thedecision
ismade:IwillupgradetoLinuxattheendofthemonthandremoveMINIX
frommywinchester,whenLinuxrunsallthesoftwareIneedandwhichcurrently
runsunderMINIX1.5withheavypatches.Iguessthismaytakeuptotwo
months.Thesearethemainreasonsformydecision:
Thereisno"current"MINIXrelease,whichcanbeusedasbasisfor
patchesandnobodyknows,when1.6.17willappear.
ThelibrarycontainsseveralbugsandfromwhatIhaveheard,thereis
noworkdoneatthem.Therewillnotbeanewcompiler,andthe16bit
usersstillhavetousebuggyACK.
1.6.17shouldoffermorePOSIX,butacompletetermiosisstillmissing.
Idoubtthatthereisstillmuchdevelopmentfor16bitusers.
IthinkIwillstopmaintainingtheMINIXsoftwarelistinafewmonths.
Anyoneoutthere,whowouldliketocontinueit?UntilLinuxruns
*perfect*onmymachine,eachupdateofOrigamiwillstillrunon16bit
MINIX.Iwillannouncewhenthelastoftheseversionsappears.
Inmyopinion,ASTisrightinhisdecisionaboutMINIX.Ireadtheflame
warandcan'tresisttosaythatIlikeMINIXthewayitis,nowwhere
thereisLinux.MINIXhassomeadvantages:
Youcanstartplayingwithitwithoutawinchester,youcaneven
compileprograms.Ididthisafewyearsago.
Itissosmall,youdon'tneedtoknowmuchtogetasmallsystemwhich
runsok.
Thereisthebook.Ok,onlyforversion1.3,butmostofitisstillvalid.
MINIXisanexampleofanonmonolithickernel.Callitamicrokernel
orahacktoovercomebraindamagedhardware:Itdemonstratesaconcept,
withitsprosandconsadocumentedconcept.
Inmyeyes,itisanicesystemforfirststepsinUNIXandsystems
programming.IlearnedmostofwhatIknowaboutUNIXwithMINIX,in
allareas,fromprogramminginCunderUNIXtosystemadministration
(andsecurityholes:)MINIXgrewwithme:1.5.xxupgrades,virtual
consoles,mail&news,textprocessing,crosscompilingetc.Nowitis
toosmallforme.Idon'tneedateachingsystemanymore,Iwouldlike
togetamorecomplicatedandfeaturefulUNIX,andthereisone:Linux.
Backintheolddays,v7wasstateoftheart.TherewasMINIXwhich
offeredmostofit.Inoneortwoyears,POSIXiswhatyouareusedto
see.Hopefully,therewillbeMINIX,offeringmostofit,withanew
book,forpeoplewhowanttorunasmallsystemtoplayandexperimentwith.
Stopflaming,MINIXandLinuxaretwodifferentsystemswithdifferent
purposes.Oneisateachingtool(andagoodoneIthink),theotheris
realUNIXforrealhackers.
Michael
From:dingbat@diku.dk(NielsSkovOlsen)
Subject:Re:1.6.17summaryandwhyIthinkASTisright.
Date:10Feb9217:33:39GMT
Organization:DepartmentofComputerScience,UofCopenhagen
http://www.oreilly.com/openbook/opensources/book/appa.html
19/20
1/7/2015
AppendixATheTanenbaumTorvaldsDebate
michael@gandalf.informatik.rwthaachen.de(MichaelHaardt)writes:
>Stopflaming,MINIXandLinuxaretwodifferentsystemswithdifferent
>purposes.Oneisateachingtool(andagoodoneIthink),theotheris
>realUNIXforrealhackers.
Hear,hear!AndnowLinuxarticlesinalt.os.linux(orcomp.os.misc
ifyoursitedon'treceivealt.*)andMinixarticleshere.
eoff(endofflamefest:)
Niels
AppendixB>
oreilly.comHome|O'ReillyBookstores|HowtoOrder|O'ReillyContacts
International|AboutO'Reilly|AffiliatedCompanies|PrivacyPolicy
2000,O'Reilly&Associates,Inc.
http://www.oreilly.com/openbook/opensources/book/appa.html
20/20