Path: nntp.gmd.de!newsserver.jvnc.net!darwin.sura.net!howland.reston.ans.net!
news.sprintlink.net!mhv.net!csbh!yoyo
From: y...@mhv.net ()
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Subject: Multi-Tasking on a PC
Date: 2 Nov 1994 03:14:54 GMT
Organization: MHVNet, the Mid Hudson Valley's Internet connection
Lines: 38
Distribution: inet
Message-ID: <39707e$k94@csbh.mhv.net>
NNTP-Posting-Host: csbh.mhv.net
Summary: RICS vs Intel, Linx vs Novell vs Solaris vs... How to choose?
X-Newsreader: TIN [version 1.2 PL2]
I though I posted this earlier, but it didn't seem to make it,
so I'll try again:
We're looking to start a small network of UNIX servers to provide
some services over the Internet. With the cost and performance of Pentiums
these days, its hard for me to justify to my small firm that we need
a mainstream RISC workstation instead of a couple of Pentiums.
I have some very general, ill-formed questions. I would
appreciate any discussion or comments that would help me (or similar others)
make a better informed decision.
1) How reliable are the current versions of UNIX on PCs?
I know lots of people use Linux et al on PCs with a single
user, but I'd like to hear from somebody who has had experience
in a situation where hundreds of processes were running
concurrently, since this is what would be happening to us.
Did the OS hold up ok, or were there problems?
2) On most architectures you get the vendor's UNIX and that's
it. On PC's you have a real choice. How to choose? Does
anybody have empirical opinions about one PC based
UNIX vs another? I am currently interested in Linux and
Solaris. Linux obviously has advantages of in that its free
and the source code is available. Solaris is SysVR4 (for
better or worse). What about other things? (I don't even
know what other things...) Performance? Support?
3) Published SPECmarks seem to show Pentiums being
comparable to or better than low end RISC workstations,
but Pentiums are a lot cheaper. Does this hold up
in practice? Yes, I know the answer is "it depends
on what you're doing", so when does it hold up and
when doesn't it?
Any advice would be greatly appreciated.
-Bruce
Path: bga.com!news.sprintlink.net!howland.reston.ans.net!swiss.ans.net!
potogold.rmii.com!rainbow!jwoznack
From: jwozn...@rmii.com (John Woznack)
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Subject: Re: Multi-Tasking on a PC
Followup-To: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Date: 4 Nov 1994 13:50:01 GMT
Organization: Rocky Mountain Internet Inc
Lines: 47
Distribution: inet
Message-ID: <39de69$69g@potogold.rmii.com>
References: <39707e$k94@csbh.mhv.net>
NNTP-Posting-Host: rainbow.rmii.com
X-Newsreader: TIN [version 1.2 PL2]
y...@mhv.net wrote:
: I though I posted this earlier, but it didn't seem to make it,
: so I'll try again:
: We're looking to start a small network of UNIX servers to provide
: some services over the Internet. With the cost and performance of Pentiums
: these days, its hard for me to justify to my small firm that we need
: a mainstream RISC workstation instead of a couple of Pentiums.
: I have some very general, ill-formed questions. I would
: appreciate any discussion or comments that would help me (or similar others)
: make a better informed decision.
: 1) How reliable are the current versions of UNIX on PCs?
: I know lots of people use Linux et al on PCs with a single
: user, but I'd like to hear from somebody who has had experience
: in a situation where hundreds of processes were running
: concurrently, since this is what would be happening to us.
: Did the OS hold up ok, or were there problems?
I'd say that UNIXWARE v1.1 is actaully very stable and reliable these days.
: 2) On most architectures you get the vendor's UNIX and that's
: it. On PC's you have a real choice. How to choose? Does
: anybody have empirical opinions about one PC based
: UNIX vs another? I am currently interested in Linux and
: Solaris. Linux obviously has advantages of in that its free
: and the source code is available. Solaris is SysVR4 (for
: better or worse). What about other things? (I don't even
: know what other things...) Performance? Support?
(My opinion) Linux is not compatable with packages written for the other
Unix OSs out there. Solaris is still buggy (so I've heard) and SCO is
just expensive. I've picked UNIXWARE myself.
: 3) Published SPECmarks seem to show Pentiums being
: comparable to or better than low end RISC workstations,
: but Pentiums are a lot cheaper. Does this hold up
: in practice? Yes, I know the answer is "it depends
: on what you're doing", so when does it hold up and
: when doesn't it?
As with any machine, the I/O seems to be the bottleneck these days. I'd
recommend getting a motherboard with VLB or PCI, and then getting an I/O
board for the VLB or PCI bus that runs your disks, serial, video, etc.
Good luck!
-JWW
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: bga.com!news.sprintlink.net!howland.reston.ans.net!EU.net!
uunet!gator!not-for-mail
From: la...@rn.com (Larry Snyder)
Subject: Re: Multi-Tasking on a PC
Message-ID: <CyszM6.3Aq@rn.com>
Organization: GatorNet, providing UUCP access to the internet!
References: <39707e$k94@csbh.mhv.net>
Distribution: inet
Date: Sat, 5 Nov 1994 16:39:42 GMT
Lines: 18
y...@mhv.net () writes:
> 3) Published SPECmarks seem to show Pentiums being
> comparable to or better than low end RISC workstations,
> but Pentiums are a lot cheaper. Does this hold up
> in practice? Yes, I know the answer is "it depends
> on what you're doing", so when does it hold up and
> when doesn't it?
I think diskIO on the Pentiums falls behind that of RISC workstations -
and any machine on the internet will be doing lots of diskIO
larry
--
Larry Snyder, System Adminstrator, GatorNet Internet Connectivity
Lake Mary (Orlando), Florida
la...@rn.com
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: bga.com!news.sprintlink.net!EU.net!uunet!gator!not-for-mail
From: la...@rn.com (Larry Snyder)
Subject: Re: Multi-Tasking on a PC
Message-ID: <Cyszor.3KD@rn.com>
Organization: GatorNet, providing UUCP access to the internet!
References: <39707e$k94@csbh.mhv.net> <39de69$69g@potogold.rmii.com>
Distribution: inet
Date: Sat, 5 Nov 1994 16:41:15 GMT
Lines: 46
jwozn...@rmii.com (John Woznack) writes:
>: 1) How reliable are the current versions of UNIX on PCs?
>: I know lots of people use Linux et al on PCs with a single
>: user, but I'd like to hear from somebody who has had experience
>: in a situation where hundreds of processes were running
>: concurrently, since this is what would be happening to us.
>: Did the OS hold up ok, or were there problems?
>I'd say that UNIXWARE v1.1 is actaully very stable and reliable these days.
I'm glad you suggest Unixware 1.1 since I am considering it as well.
>: 2) On most architectures you get the vendor's UNIX and that's
>: it. On PC's you have a real choice. How to choose? Does
>: anybody have empirical opinions about one PC based
>: UNIX vs another? I am currently interested in Linux and
>: Solaris. Linux obviously has advantages of in that its free
>: and the source code is available. Solaris is SysVR4 (for
>: better or worse). What about other things? (I don't even
>: know what other things...) Performance? Support?
>(My opinion) Linux is not compatable with packages written for the other
>Unix OSs out there. Solaris is still buggy (so I've heard) and SCO is
>just expensive. I've picked UNIXWARE myself.
I agree. Plus SCO is based on 3.2 Unix and has limitations..
>: 3) Published SPECmarks seem to show Pentiums being
>: comparable to or better than low end RISC workstations,
>: but Pentiums are a lot cheaper. Does this hold up
>: in practice? Yes, I know the answer is "it depends
>: on what you're doing", so when does it hold up and
>: when doesn't it?
>As with any machine, the I/O seems to be the bottleneck these days. I'd
>recommend getting a motherboard with VLB or PCI, and then getting an I/O
>board for the VLB or PCI bus that runs your disks, serial, video, etc.
who's board and what buss are you using for your Unixware Server?
larry
--
Larry Snyder, System Adminstrator, GatorNet Internet Connectivity
Lake Mary (Orlando), Florida
la...@rn.com
Path: bga.com!news.sprintlink.net!howland.reston.ans.net!cs.utexas.edu!
utnut!utgpu!attcan!telly!evan
From: e...@telly.on.ca (Evan Leibovitch)
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Subject: Re: Multi-Tasking on a PC
Message-ID: <CytuH9.AxB@telly.on.ca>
Date: 6 Nov 94 03:46:21 GMT
References: <39707e$k94@csbh.mhv.net> <CyszM6.3Aq@rn.com>
Organization: Sound Software Ltd., Brampton, Ontario
Lines: 19
In article <CyszM6....@rn.com>, Larry Snyder <la...@rn.com> wrote:
>I think diskIO on the Pentiums falls behind that of RISC workstations -
>and any machine on the internet will be doing lots of diskIO
Is that *necessarily* still the case with the maturing of PCI, and
UnixWare accessibility to fast/wide SCSI drives?
Is a good PCI bus-mastering controller like the Adaptec or Buslogic,
matched with a driver that exploits its power, really that inferior
technically to the RISC equivalents?
It's only a matter of time before we start to see a full range of
caching PCI disk controllers similar to the choices now for EISA.
--
Evan Leibovitch, Sound Software Ltd., located in beautiful Brampton, Ontario
Novell Unix Master Reseller / e...@telly.on.ca / (905) 452-0504
PCMCIA: People Can't Memorize Computer Industry Acronyms
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!agate!darkstar.UCSC.EDU!
cats.ucsc.edu!philb
From: ph...@cats.ucsc.edu (Philip Brown)
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Subject: Re: Multi-Tasking on a PC
Date: 7 Nov 1994 01:41:58 GMT
Organization: University of California, Santa Cruz
Lines: 22
Distribution: inet
Message-ID: <39k0l6$f2n@darkstar.UCSC.EDU>
References: <39707e$k94@csbh.mhv.net> <39de69$69g@potogold.rmii.com>
NNTP-Posting-Host: am.ucsc.edu
In article <39de69$...@potogold.rmii.com>,
John Woznack <jwozn...@rmii.com> wrote:
>
>I'd say that UNIXWARE v1.1 is actaully very stable and reliable these days.
oh sure. (apart from the nfs). its just a pain to work with.
for example, try getting gcc to compile. and no gdb support.
I wish to hell our site would have gone all solaris instead of mixed
solaris & unixware.
>
>(My opinion) Linux is not compatable with packages written for the other
>Unix OSs out there. Solaris is still buggy (so I've heard)
not more than unixware.
--
----------------------------------------------------------------------
Philip Brown, author of "kdrill", and "xmandel"
ph...@cats.ucsc.edu ph...@csua.berkeley.edu
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!news.moneng.mei.com!
hookup!news.mathworks.com!uhog.mit.edu!news.mtholyoke.edu!nntp.et.byu.edu!
news.provo.novell.com!uel!stubai!msohnius
From: msohn...@novell.co.uk (Martin Sohnius)
Subject: Re: Multi-Tasking on a PC
Sender: n...@novell.co.uk
Message-ID: <Cyx5Jn.70J@novell.co.uk>
Date: Mon, 7 Nov 1994 22:38:11 GMT
Distribution: inet
References: <39707e$k94@csbh.mhv.net> <39de69$69g@potogold.rmii.com>
<39k0l6$f2n@darkstar.UCSC.EDU>
Organization: Novell Europe
X-Newsreader: TIN [version 1.2 PL2]
Followup-To: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Lines: 25
Philip Brown (ph...@cats.ucsc.edu) wrote:
: In article <39de69$...@potogold.rmii.com>,
: John Woznack <jwozn...@rmii.com> wrote:
: >
: >I'd say that UNIXWARE v1.1 is actaully very stable and reliable these days.
: oh sure. (apart from the nfs). its just a pain to work with.
^^^^^^^^^^^^^^^^^^
why do you say this?
: for example, try getting gcc to compile. and no gdb support.
"Getting gcc to compile" on a bog-standard SVR4 compilation system sounds
more like a problem with gcc to me, than with UnixWare. Besides which,
compiled versions are available for download. Use them to compile later
versions, if the AT&T compiler isn't good enough for you.
--
+----------------------------------+
Martin Sohnius | "If you can't be funny, |
Novell Labs Europe | at least be interesting." |
Bracknell, England | - Harold W. Ross |
+44-1344-724031 +----------------------------------+
(I speak for myself, not for Novell or anyone else.)
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!cs.utexas.edu!
uunet!news.widomaker.com!escape!shendrix
From: shend...@escape.widomaker.com (Shannon Hendrix)
Subject: Re: Multi-Tasking on a PC
Distribution: inet
Organization: HNN UNIX Network
Message-ID: <1994Nov6.170129.5275@escape.widomaker.com>
References: <39707e$k94@csbh.mhv.net> <CyszM6.3Aq@rn.com>
Date: Sun, 6 Nov 1994 17:01:29 GMT
Lines: 19
la...@rn.com (Larry Snyder) writes:
>y...@mhv.net () writes:
>I think diskIO on the Pentiums falls behind that of RISC workstations -
>and any machine on the internet will be doing lots of diskIO
Not unless you are running very fast hard drives and special disk
controllers (like wuarchive's DEC system). Given that Linux on a PC
can run SCSI drives at near maximum throughput I don't see how a
workstation could do better. A VME-bus system could probably handle
more drives at a given speed than PCI but you'd have to have quite a
few and then I'd say you would be overloading all but the fastest
workstations.
--
csh
---------------------------------------------------------------------------
shend...@escape.widomaker.com | Linux... that's it for the moment
-----------------------------------+
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!torn!news.unb.ca!
upei.ca!peinet.pe.ca!krman!krman
From: kr...@peinet.pe.ca (Kevin MacRae)
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Subject: Re: Multi-Tasking on a PC
Date: 9 Nov 1994 14:34:42 GMT
Organization: PEINet Inc, Charlottetown, PEI Canada
Lines: 27
Distribution: inet
Message-ID: <39qmm2$2h6@bud.peinet.pe.ca>
References: <39707e$k94@csbh.mhv.net> <CyszM6.3Aq@rn.com>
<1994Nov6.170129.5275@escape.widomaker.com>
NNTP-Posting-Host: 198.167.2.9
In article <1994Nov6.170129.5...@escape.widomaker.com>,
shend...@escape.widomaker.com (Shannon Hendrix) writes:
|> la...@rn.com (Larry Snyder) writes:
|>
|> >y...@mhv.net () writes:
|>
|> >I think diskIO on the Pentiums falls behind that of RISC workstations -
|> >and any machine on the internet will be doing lots of diskIO
|>
|> Not unless you are running very fast hard drives and special disk
|> controllers (like wuarchive's DEC system). Given that Linux on a PC
|> can run SCSI drives at near maximum throughput I don't see how a
|> workstation could do better. A VME-bus system could probably handle
|> more drives at a given speed than PCI but you'd have to have quite a
|> few and then I'd say you would be overloading all but the fastest
|> workstations.
The main problem with PC tech. is that people are stuffing it with DOS hardware.
.i.e. IDE drives, 8550/16550 com-ports, interrupt driven printer ports, etc.
Installing good hardware can match most systems $ for $ Which is what it is
all about. If you spend $10k on a RISC, spend $10k on a Pen.
If you spend $20k on a RISC, spend $20k on a Pen. And so on.
*******************************************************************************
* Kevin MacRae Phone (902) 566-3198 *
* Owner/Operator FAX (902) 566-3423 *
* K & R Management Internet kr...@peinet.pe.ca *
*******************************************************************************
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!cs.utexas.edu!
uunet!psinntp!darkstar!elt
From: e...@irony.com (Ed Taychert)
Subject: Re: Multi-Tasking on a PC
Message-ID: <Cz21MJ.1tu@irony.com>
Organization: Irony Games
References: <39707e$k94@csbh.mhv.net> <CyszM6.3Aq@rn.com> <CytuH9.AxB@telly.on.ca>
Date: Thu, 10 Nov 1994 14:01:31 GMT
Lines: 30
In article <CytuH9....@telly.on.ca>, Evan Leibovitch <e...@telly.on.ca> wrote:
>In article <CyszM6....@rn.com>, Larry Snyder <la...@rn.com> wrote:
>
>>I think diskIO on the Pentiums falls behind that of RISC workstations -
>>and any machine on the internet will be doing lots of diskIO
>
>Is that *necessarily* still the case with the maturing of PCI, and
>UnixWare accessibility to fast/wide SCSI drives?
>
>Is a good PCI bus-mastering controller like the Adaptec or Buslogic,
>matched with a driver that exploits its power, really that inferior
>technically to the RISC equivalents?
>
>It's only a matter of time before we start to see a full range of
>caching PCI disk controllers similar to the choices now for EISA.
I think most folks' internet connections are slower than their disks.
I agree that PC unix vendors have to get on the fast/wide bandwagon
though. With multiple disks on my Solaris x86 box, I can see
sustained SCSI usage of 5Mb/sec, the limit (I think!) of async SCSI.
(I have an Adaptec vlb controller) I think the drivers are slowing
me down, not the hardware ...
- Ed.
--
-------------------------------------------------------------------------
Ed Taychert | No plants or animals were killed to create this message.
e...@irony.com | It is composed entirely of recycled computer bits.
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!news.moneng.mei.com!
uwm.edu!lll-winken.llnl.gov!koriel!newsworthy.West.Sun.COM!
abyss.West.Sun.COM!astro!richard
From: rich...@astro.West.Sun.COM (Richard Mathews)
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Subject: Re: Multi-Tasking on a PC
Date: 10 Nov 1994 22:16:58 GMT
Organization: Sunsoft Inc., Los Angeles Site, Ca.
Lines: 22
Distribution: inet
Message-ID: <39u64q$7k6@abyss.West.Sun.COM>
References: <39707e$k94@csbh.mhv.net> <39de69$69g@potogold.rmii.com>
<Cyszor.3KD@rn.com> <Cywu9x.1HE@rheged.dircon.co.uk>
NNTP-Posting-Host: astro.west.sun.com
X-Newsreader: NN version 6.5.0 (NOV)
In article <Cyszor....@rn.com>, Larry Snyder <la...@rn.com> wrote:
>>(My opinion) Linux is not compatable with packages written for the other
>>Unix OSs out there. Solaris is still buggy (so I've heard) and SCO is
>>just expensive. I've picked UNIXWARE myself.
>
>I agree. Plus SCO is based on 3.2 Unix and has limitations..
I disagree with respect to Solaris. Early releases of Solaris, including
2.1 for x86, were buggy. Solaris 2.4 is quite solid. I am running 2.4
for x86 on my desktop. I have not bothered to install a single patch.
The machine has been running without any problems at all for 3.5 months.
There are still problems (like "hardware configuration is a bitch"), but
Solaris does what it claims to do very well. Don't say "Solaris is still
buggy" unless you have experiences with 2.4.
Note: I am biased.
Richard M. Mathews F oster
E stonian-Latvian-Lithuanian
rich...@West.Sun.COM I ndependence and
F reedom!
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!gatech!
newsxfer.itd.umich.edu!zip.eecs.umich.edu!yeshua.marcam.com!
news.mathworks.com!uhog.mit.edu!news.mtholyoke.edu!nntp.et.byu.edu!
news.provo.novell.com!uel!stubai!msohnius
From: msohn...@novell.co.uk (Martin Sohnius)
Subject: Re: Multi-Tasking on a PC
Sender: n...@novell.co.uk
Message-ID: <Cz2EEI.2nK@novell.co.uk>
Date: Thu, 10 Nov 1994 18:37:30 GMT
Distribution: inet
References: <39707e$k94@csbh.mhv.net> <39de69$69g@potogold.rmii.com>
<39k0l6$f2n@darkstar.UCSC.EDU> <Cyx5Jn.70J@novell.co.uk>
<39ofas$bf4@darkstar.UCSC.EDU>
Organization: Novell Europe
X-Newsreader: TIN [version 1.2 PL2]
Followup-To: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Lines: 43
Philip Brown (ph...@cats.ucsc.edu) wrote:
: In article <Cyx5Jn....@novell.co.uk>,
: Martin Sohnius <msohn...@novell.co.uk> wrote:
: >"Getting gcc to compile" on a bog-standard SVR4 compilation system sounds
: >more like a problem with gcc to me, than with UnixWare.
: >compiled versions are available for download. Use them to compile later
: >versions, if the AT&T compiler isn't good enough for you.
: Wel, the problem is that nothingg GNU supports it.
I can't parse that sentence. However, whatever it may try to say,
the following is direct output from my system:
$ gcc -Wall -c tst.c
tst.c:3: warning: return-type defaults to `int'
tst.c: In function `main':
tst.c:18: warning: control reaches end of non-void function
$
(obviously, I didn't take too much care writing this little program...)
Now, if gcc supported the same -V option as UnixWare's 'cc', I could
even tell you the version :->)
: So there's no way I'll be aable to "compile later versions."
Why not?
: (and no the at&t compiler isn't good enough :-)
Must be because the GNU folks know C better than D. Ritchie, then...
(yes, yes, I know: it's been a long time since he had anything to
do with the current acomp :-).
--
+----------------------------------+
Martin Sohnius | "If you can't be funny, |
Novell Labs Europe | at least be interesting." |
Bracknell, England | - Harold W. Ross |
+44-1344-724031 +----------------------------------+
(I speak for myself, not for Novell or anyone else.)
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!swrinde!pipex!
uknet!info!iialan
From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: Multi-Tasking on a PC
Message-ID: <Cz9DIH.16r@info.swan.ac.uk>
Sender: n...@info.swan.ac.uk
Nntp-Posting-Host: iifeak.swan.ac.uk
Organization: Institute For Industrial Information Technology
References: <Cyszor.3KD@rn.com> <Cywu9x.1HE@rheged.dircon.co.uk>
<1994Nov8.043215.15815@kf8nh.wariat.org>
Distribution: inet
Date: Mon, 14 Nov 1994 13:01:28 GMT
Lines: 13
In article <1994Nov8.043215.15...@kf8nh.wariat.org> b...@kf8nh.wariat.org
(Brandon S. Allbery) writes:
>Several of us have tried, but many of the Linux developers are BSD-worshippers
>at heart and won't have anything to do with V.4.
And may of us are ex V.3+streams/V.4 people who still have nightmares about
it 8).
Alan
--
..-----------,,----------------------------,,----------------------------,,
// Alan Cox // iia...@www.linux.org.uk // GW4PTS@GB7SWN.#45.GBR.EU //
``----------'`----------------------------'`----------------------------''
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!pipex!uknet!info!iialan
From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: Multi-Tasking on a PC
Message-ID: <CzB497.FBK@info.swan.ac.uk>
Sender: n...@info.swan.ac.uk
Nntp-Posting-Host: iifeak.swan.ac.uk
Organization: Institute For Industrial Information Technology
References: <CyszM6.3Aq@rn.com> <CytuH9.AxB@telly.on.ca> <Cz21MJ.1tu@irony.com>
Date: Tue, 15 Nov 1994 11:36:43 GMT
Lines: 21
In article <Cz21MJ....@irony.com> e...@irony.com (Ed Taychert) writes:
>I agree that PC unix vendors have to get on the fast/wide bandwagon
>though. With multiple disks on my Solaris x86 box, I can see
>sustained SCSI usage of 5Mb/sec, the limit (I think!) of async SCSI.
>(I have an Adaptec vlb controller) I think the drivers are slowing
>me down, not the hardware ...
PCI bus with multiple SCSI controllers is always a good idea anyway
if you can afford it and your OS allows you to do it (and the NCR PCI
controllers are pretty cheap). A couple for the disks, one for the swap disk
and one to hang the tap/cdroms off should do most people nicely 8).
PCI performance certainly seems to be workstation or better. Certainly
better than some of the low end workstations which often do have a fairly
slow bus.
Alan
--
..-----------,,----------------------------,,----------------------------,,
// Alan Cox // iia...@www.linux.org.uk // GW4PTS@GB7SWN.#45.GBR.EU //
``----------'`----------------------------'`----------------------------''
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!pipex!uknet!info!iialan
From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: Multi-Tasking on a PC
Message-ID: <CzB4Ft.FEB@info.swan.ac.uk>
Sender: n...@info.swan.ac.uk
Nntp-Posting-Host: iifeak.swan.ac.uk
Organization: Institute For Industrial Information Technology
References: <Cyx5Jn.70J@novell.co.uk> <39ofas$bf4@darkstar.UCSC.EDU>
<Cz2EEI.2nK@novell.co.uk>
Distribution: inet
Date: Tue, 15 Nov 1994 11:40:41 GMT
Lines: 24
In article <Cz2EEI....@novell.co.uk> msohn...@novell.co.uk (Martin Sohnius)
writes:
>Now, if gcc supported the same -V option as UnixWare's 'cc', I could
>even tell you the version :->)
It does gcc -v. And then there is gcc -V x.y.z if you want to compile with
an older backend you also have installed. Wonderful for regression testing
jobs.
>Must be because the GNU folks know C better than D. Ritchie, then...
gcc -ansi is sufficiently accurate it shows up bugs in some ANSI compliance
suites. More to the point I trust gcc2.5.8 to get it right. I have no
confidence in most other compilers, especially as I do a lot of driver
level work where things like volatile _HAVE_ to work properly.
Gcc in GNU extensions mode is a different matter and I'm not too keen on
people who write in gcc not Ansi C.
Alan
--
..-----------,,----------------------------,,----------------------------,,
// Alan Cox // iia...@www.linux.org.uk // GW4PTS@GB7SWN.#45.GBR.EU //
``----------'`----------------------------'`----------------------------''
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!pipex!uknet!info!iialan
From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: Multi-Tasking on a PC
Message-ID: <CzB4Mp.FHC@info.swan.ac.uk>
Sender: n...@info.swan.ac.uk
Nntp-Posting-Host: iifeak.swan.ac.uk
Organization: Institute For Industrial Information Technology
References: <Cyszor.3KD@rn.com> <Cywu9x.1HE@rheged.dircon.co.uk>
<39u64q$7k6@abyss.West.Sun.COM>
Distribution: inet
Date: Tue, 15 Nov 1994 11:44:48 GMT
Lines: 21
In article <39u64q$...@abyss.West.Sun.COM> rich...@astro.West.Sun.COM
(Richard Mathews) writes:
>Solaris does what it claims to do very well. Don't say "Solaris is still
>buggy" unless you have experiences with 2.4.
Ok Solaris is no more broken than anyone else now (and the networking speed
is impressive even if it's sticking to specifications is as hopeless as
BSD). It still doesn't seem to report ICMP responses to UDP sendto()'s to
the application layer (as RFC1122 requires), and from reports though I've
not tried this myself the stream buffer monster still gives it bad
indigestion (Thats a program that opens 256 UDP sockets and fills them with
1 byte sendto's until the machine seems out of buffers then forks and the
child starts the task again while the master sits on tons of buffers).
Solaris does now seem to be getting to the 'You have to try and break it'
stage which is ahead of a lot of people.
Alan
--
..-----------,,----------------------------,,----------------------------,,
// Alan Cox // iia...@www.linux.org.uk // GW4PTS@GB7SWN.#45.GBR.EU //
``----------'`----------------------------'`----------------------------''
Newsgroups: comp.unix.solaris,comp.os.linux.misc,comp.unix.unixware,
comp.unix.xenix.sco,comp.unix.pc-clone.32bit
Path: nntp.gmd.de!xlink.net!howland.reston.ans.net!pipex!uknet!info!iialan
From: iia...@iifeak.swan.ac.uk (Alan Cox)
Subject: Re: Multi-Tasking on a PC
Message-ID: <CzF0rv.2s9@info.swan.ac.uk>
Sender: n...@info.swan.ac.uk
Nntp-Posting-Host: iifeak.swan.ac.uk
Organization: Institute For Industrial Information Technology
References: <CyszM6.3Aq@rn.com> <1994Nov6.170129.5275@escape.widomaker.com>
<39qmm2$2h6@bud.peinet.pe.ca>
Distribution: inet
Date: Thu, 17 Nov 1994 14:11:54 GMT
Lines: 26
In article <39qmm2$...@bud.peinet.pe.ca> kr...@peinet.pe.ca (Kevin MacRae)
writes:
>The main problem with PC tech. is that people are stuffing it with DOS hardware.
>.i.e. IDE drives, 8550/16550 com-ports, interrupt driven printer ports, etc.
>Installing good hardware can match most systems $ for $ Which is what it is
>all about. If you spend $10k on a RISC, spend $10k on a Pen.
> If you spend $20k on a RISC, spend $20k on a Pen. And so on.
Ummm: PC hardware of the same quality is fundamentally cheaper due to market
size. Sun floppy drives are 130 UKP here, PC ones 25 UKP and up to say 30UKP
for a good respected brand.
Also if a mouse a 16450 com port is more than perfect, and for one modem a
16550A is superb. The printer is more than adequate for one printer and so
on. If I wanted a 512 line modem server I'd buy a big terminal server (or
three).
For the same price you should be getting better hardware, more
expandability, more options (the PC architecture is open), cheaper and
actually purchasable spares ['Sorry you don't have a maintenance contract
we can't supply you with an xxxx'].
Alan
--
..-----------,,----------------------------,,----------------------------,,
// Alan Cox // iia...@www.linux.org.uk // GW4PTS@GB7SWN.#45.GBR.EU //
``----------'`----------------------------'`----------------------------''