Subject: Linux-Misc Digest #559
From: Digestifier <Linux-Misc-Request@senator-bedfellow.MIT.EDU>
To: Linux-Misc@senator-bedfellow.MIT.EDU
Reply-To: Linux-Misc@senator-bedfellow.MIT.EDU
Date:     Sat, 6 Aug 94 07:13:10 EDT

Linux-Misc Digest #559, Volume #2                 Sat, 6 Aug 94 07:13:10 EDT

Contents:
  utility for creating video modes for X? (cloister bell)
  Re: DOOM For X? (Gregory Scott Murray)
  Re: unzip 5.11 released by Info-zip (Cave Newt)
  Re: Linux vs. CP/M ? (Farrell McGovern)
  Re: WABI vs. SoftWindows? (Colin A. Frank)
  Re: source of TCP/IP (was I hope this wont ignite a major flame ...) (Larry Doolittle)
  Re: I hope this won't ignite a major flame war, but I've got to know! (Callum Gibson)
  Re: Reliable NFS server, Logical File System? (Donald Becker)
  Re: Term and gopher - now.... (Elan Feingold)
  Re: WABI vs. SoftWindows? (Alan Braggins)
  Re: Just Computers advertising practices (Tim Smith)
  Re: WABI vs. SoftWindows? (Tim Cutts)
  Re: how to use term 2.04? (Patrick Reijnen)
  `mkisofs' ? (Ajay Shah)
  Re: VL-Bus SCSI adaptors (Christopher M. May)
  linux-1.1.39 doesn't wanna compile w/TMC850 Driver (Nevin Cheung)
  dialin for X
  Re: STREAMS  (was I hope this wont ignite ...) (Terry Lambert)
  Re: Usefulness of BSD/Linux Source Knowledge (was BSD vs. LINUX) (Terry Lambert)

----------------------------------------------------------------------------

From: cloister@u.washington.edu (cloister bell)
Subject: utility for creating video modes for X?
Date: 6 Aug 1994 01:13:01 GMT

i read through VideoModes.doc, and worked through the steps for creating a
video mode, and well, i was pleased to find that the results actually worked.
but it occurs to me that someone has to have written a utility to do this
automatically.  something that would take a dcf value and a hsf value and give
you back a line that you can stick in your Xconfig file.  has anyone seen
something like this?
-- 
+-------------------------------------------------+---------------------------+
|tactical nuclear sdi stealth nsafood signature.  | cloister@u.washington.edu |
+-------------------------------------------------+---------------------------+

------------------------------

From: murraygs@newton.ccs.tuns.ca (Gregory Scott Murray)
Subject: Re: DOOM For X?
Date: Wed, 3 Aug 1994 03:11:53 GMT

In article <31fm35$ck4@harbinger.cc.monash.edu.au> ckup1@aurora.cc.monash.edu.au (Christopher Kuperman) writes:
>
>       Would I be wrong in saying that DOOM for MS-DOS was
>       solely constructed and designed to utilize and exploit
>       very low-level hardware features. + they have always tried
>       to squeeze every last MIP out of the poor 386/486 architecture.

Yes! :-) In several interviews/posts by the developers, they have
stressed that the majority of DOOM is written in ANSI C.  Only a
couple of texture mapping routines, and I think some joystick code are
written in assembly.  As well, it was developed under NextStep, and a
version has been available from the start for that platform.  The port
to the Jaguar is nearing completion, and the SGI/Irix port has just
been released, with the Linux port following soon.

>       To move DOOM to OS/2, X, or even Chicago would devistate the
>       performance of it. I find that DOOM running on a 486-66 with
>       a VLB SVGA card -just- approaches acceptable performance. 
>       Shall we guess as to what one might need while running doom
>       under Win-Doze.?

So far, the posts in the alt.games.doom newsgroup seem quite positive
about the SGI port, so it's looking good for the Linux version.  IBM
are doing the port to OS/2 themselves, and not much has been said
about it.  However, from the rumor mill, using the new WinG toolkit,
Microsoft and id have been showing DOOM running on 486-based PC's in a
window at acceptable speeds.  Since some form of the WinG package is
going to be part of Chicago, it seems plausible that Windows DOOM is
not too far away.

Scott


--
===============================================================================
      "Good, bad ... I'm the guy with the gun." - Ash, "Army of Darkness"
===============================================================================
                Scott Murray : murraygs@newton.ccs.tuns.ca 
    3rd Year Computer Science Student | Technical University of Nova Scotia
===============================================================================

------------------------------

From: roe2@ellis.uchicago.edu (Cave Newt)
Subject: Re: unzip 5.11 released by Info-zip
Reply-To: roe2@midway.uchicago.edu
Date: Sat, 6 Aug 1994 01:11:04 GMT

gelinas@VAX1.cmr.ca (Dr. Jacques Gelinas) writes:

>The newest version of  unzip  was released yesterday by the Info-zip group.
>It is of course available at xxxxxxxxxxxxxxxxxx, the info-zip home site.

...and we'd appreciate it if people would not publicize (or use) this
site--it's the personal workstation of one of the developers.

ftp.uu.net:/pub/archiving/zip is the official primary mirror site for
public releases, and it's much better able to handle a large ftp load.

>Note that zip seem more effective than compress on the unzip archive (:).
>-r--r--r--  1 ftp       513509 Aug  2 23:18 unzip511.tar.Z
>-r--r--r--  1 ftp       397944 Aug  2 23:18 unzip511.zip

For fairness, however, you should compare with tar+gzip, which compresses
much better than zip (due to its effective use of "unit packing").  Of 
course, you have to decompress the whole archive to get a listing or to
extract a single file...

>-rw-r--r--  1 zipper    106236 Jul 27 13:54 novbug1.zip
>-rw-r--r--  1 zipper     26404 Jul 27 13:54 novbug2.zip

These are actually demonstrations of a bug in PKZIP. :-)  Zip functions
correctly on Novell drives.  See the ToDo file for a list of what's next
on the slate.

>  gzip -cd linux-1.37.tar.gz | tar tvf - > /dev/null    11.90 sec (The winner)
>  funzip   linux-1.37.tar.gz | tar tvf - > /dev/null    13.98 sec

fUnZip should improve in the next release when we switch to a "new"
version of inflate (which has actually been around a long time, but
never got incorporated for various historical reasons).  I'm pretty
sure gunzip already uses the faster version.

>New with the 511 release is a stub, unzipsfx (40k), to transform a zip archive
>into a self decompressing archive. Here are the comments from the man page:

Note that there is a newly discovered bug in this (oh, the embarrassment),
and as a result there will be a 5.12 release in roughly 10 days.  The bug,
btw, involves an off-by-one pointer and results in the loss of the first
filespec (e.g., "foosfx '*.[ch]' \*.doc" only extracts *.doc).  The fix is
trivial; there's a "wildzipfn = *argv++" around line 693, I believe (this
is from memory), and it should be surrounded by an "#ifndef SFX".

As a personal point of interest, I'd be curious to know whether folks
would like/dislike/not care to have UnZip binaries distributed in a
self-extracting archive instead of a gzip'd tarfile.  I don't really
know the issues regarding shared library versions, iBCS2 (or whatever),
etc.  Responses via e-mail (to me or zip-bugs) are fine.

>The unzip511.zip file unzips into the current directory.

The tarfile creates an unzip-5.11 directory, however.

>the existing unzip (5.0pl1) in /usr/bin (Slackware 2.0)

They're still using 5.0p1?  I must have words with them...

>Jacques Ge'linas, Ph.D., Maths,           INTERNET: gelinas@cmr.ca

Thanks for the local announcement.  I would have gotten here eventually
(yes, three of the core authors use Linux), but I figured Linux users
can fend for themselves a little better than DOS or even OS/2 folks. :-)

Greg Roelofs
UnZip maintainer
(running Linux, OS/2, Doom :-) )

------------------------------

From: ai474@FreeNet.Carleton.CA (Farrell McGovern)
Subject: Re: Linux vs. CP/M ?
Reply-To: ai474@FreeNet.Carleton.CA (Farrell McGovern)
Date: Wed, 3 Aug 1994 17:48:55 GMT



        Ghods! Did I get a good laugh when I saw this subject line!!!

In a previous article, grante@reddwarf.rosemount.com (Grant Edwards) says:

>Let's be fair, should we compare to CP/M 1.7, 2.2, or 3.0?  I spent
>most of my time with CP/M 1.7, so I'll address Linux's faults with
>respect to that.

        I used the old Xerox 820's and Apple ][+ Z80 cards (which I
assembled myself)

>Can Linux run on an Intel MDS-800 "blue-box?"  I didn't think so --
>clearly a gross deficiency.  And what about the lack of FCBs for file
>I/O?  I'm sure you guys have some lame excuse. Does Linux have PIP?
>No -- yet another shortcoming.  Building a Linux kernel isn't nearly
>as interesting as doing a SYSGEN and then running DDT to patch in the
>latest CBIOS.  And how about language support?  Does Linux have a port
>of Eubanks' CBASIC compiler?  I didn't think so!

        I remember patching Wordstar so that I could get compressed and
double width charactors out of the Epson MX-80 printer I had...ah...those
were the days...

ttyl
     Farrell
-- 
Hail Eris!                                         All Hail Discordia!
            *->  Current Magicknet Moderator  <-*
Sysop Solsbury Hill BBS                            PODSnet 93:9631/523
Ask me about PODSnet, Paganism, Eris, Wicca...and I might even answer!

------------------------------

From: cfrank@zia (Colin A. Frank)
Crossposted-To: comp.sys.sun.apps,comp.windows.x,comp.windows.x.apps
Subject: Re: WABI vs. SoftWindows?
Date: 3 Aug 1994 18:02:24 GMT

In article 6ik@lyra.csx.cam.ac.uk, tjrc1@cus.cam.ac.uk (Tim Cutts) writes:
>>>job control, 
>
>>Why on earth is job control essential?

Hmmm!!   I don't know??  Especially when you could just hit the reset
button or hit the power switch for a second or two.  That is the most
effective way to control all jobs.

>
>How else do you manage a multitasking system?  Under OS/2, clicking
>both mouse buttons gets me a list of running processes.  Click right
>button on it and I get a menu of options such as show, close, and so

Agreed!

>on.  Rather simpler than pissing about with ps, kill, fg, bg and
>process ids.
>

Cool => Warm => Hot =>  <<<~~~ FLAME :-( ~~~>>>

Oh neat!  This is great for an nintendo-like OS that assumes that the
only user is the one *physically* loacted at the console with joystick (I
mean mouse) in hand.  Some systems are designed to be multiuser and
provide facilities to administer the system remotely. Non-interactive
use is also a plus.

Let's test your neat little system: From home, call up your OS/2 system
at work and terminate any proccess owned by Jack or Jill.
 1. Can multiple users login to OS/2 simultaneously?
 2. Do they make mouse cables that long? :-)

=============================================================================
Colin... H Voice: 510-2**-2***  amdahl.com!zia!cfrank H Data: 510-2**-5***
 \|/     W Voice: 510-9**-0***  cfrank@sybase.com
  *==>   Water under the bridge comes back at high tide.
=============================================================================



------------------------------

From: doolitt@recycle.cebaf.gov (Larry Doolittle)
Subject: Re: source of TCP/IP (was I hope this wont ignite a major flame ...)
Reply-To: doolittle@cebaf.gov
Date: Fri, 5 Aug 1994 20:39:59 GMT

Warner Losh (imp@boulder.parcplace.com) wrote:

: There's a big difference between code that humans can read, and code
: that requires mutants to read.  The code that Vernon quoted is an
: example of code that mutants read. 
  [snip]
: Linux does a great job at what it does, but some parts of it are icky
: to look at.  Kinda like the plumbing in my house.

So quit griping and post some patches!

         - Larry Doolittle   doolittle@cebaf.gov

------------------------------

Crossposted-To: comp.os.386bsd.misc
From: callum@frost.bain.oz.au (Callum Gibson)
Subject: Re: I hope this won't ignite a major flame war, but I've got to know!
Date: Tue, 2 Aug 1994 04:17:11 GMT

Nate Williams (nate@bsd.coe.montana.edu) wrote:
> None of the 'real development' is being done there.  (I've been reading
> the Linux newsgroups since their ineption)
                                   ^^^^^^^^

Freudian slip there, Nate? :-)

-- 
Callum Gibson                                             callum@bain.oz.au
Fixed Income Division, DB Bain & Co.                          61 2 258 1620

------------------------------

From: becker@cesdis.gsfc.nasa.gov (Donald Becker)
Subject: Re: Reliable NFS server, Logical File System?
Date: 5 Aug 1994 22:55:39 -0400

In article <KJETILHO.94Aug5010142@mnemosyne.uio.no>,
Kjetil Torgrim Homme <kjetilho@mnemosyne.uio.no> wrote:
>+--- Taner Ozcelik:
>| Currently I have an application running on DEC-5000 that generates
>| about 50000 bytes/sec worth of data, contineously.
>| [...]
>| VERY IMP is stabilty to provide a fast and reliable filesystem. I
>| guess my IntelEtherExpress should not be a problem as it gives me
>| ~300Kb/sec ftp transferrate.
>
>Hmm, FTP is pretty much best-case for TCP/IP through-put. Linux

Actually opening a TCP pipe with the 'ttcp' program is slightly better.
More importantly 'ttcp' always opens a raw pipe, while 'ftp' programs often
default to "text" mode which can dramatically drop the transfer rate.  Even
in binary mode 'ftp' might be testing the file system more than the network.

>kernels prior to 1.1.27 or thereabouts has very poor NFS performance,
>like 40-50kB/s. That's with a 3c509 card which I think is quite a bit
>faster than the EtherExpress, it has a peak rate of 7-900kB/s.

Generally the lowest NFS rates are writing to a Sun host.  Reading is much
faster than writing on most servers, and Suns only work well with 8K writes.  
Auspex servers have very good performance doing anything:->.


-- 
Donald Becker                                     becker@cesdis.gsfc.nasa.gov
USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center,  Greenbelt, MD.  20771
301-286-0882         http://cesdis.gsfc.nasa.gov/pub/people/becker/whoiam.html

------------------------------

From: feingold@avette.zko.dec.com (Elan Feingold)
Crossposted-To: comp.os.linux.help
Subject: Re: Term and gopher - now....
Date: 3 Aug 1994 18:26:17 GMT
Reply-To: feingold@avette.zko.dec.com (Elan Feingold)


> Well, if you have ftptool running with term, I'm sure there are a few
> people who would like to get it (there was a discussion about this a
> while ago and a few people would get this Segmentation Fault error
> at run time).

I got that until I relinked with static XView libs, then it worked fine.

Elan

--
===============================================================================
|| Elan Feingold (Cornell '94)    || "Two of the most famous products of     ||
|| Software Engineer II           ||  Berkeley are LSD and Unix. I don't     ||
|| Digital Equipment Corporation  ||  think that is a coincidence."          ||
|| Work: 603.881.1115             ||                       - Anonymous       ||
===============================================================================

------------------------------

Crossposted-To: comp.sys.sun.apps,comp.windows.x,comp.windows.x.apps
From: armb@setanta.demon.co.uk (Alan Braggins)
Subject: Re: WABI vs. SoftWindows?
Date: Fri, 5 Aug 1994 10:36:20 GMT

In article <31nrg8$4f7@lyra.csx.cam.ac.uk> tjrc1@cus.cam.ac.uk (Tim Cutts) writes:
   The Cambridge University Central Unix Service is a four-machine NFS
   connected monster running SunOS 4.1.3 on three machines and Solaris on
   the fourth.  We have something on the order of 10,000 users across the
   university.  Each machine currently has about 70 users logged on.

And running Windows on those machines would be simpler? This is hardly
relevant to discussion of their relative merits as desktop systems.
Windows won't support 70 users at once.

   >"Hello, I just screwed up my config.sys and autoexec.bat for the 1,000th
   > time.  Since we have no security under MS-"DOS", would you please WALK
   > doen here and fix it for me again"

   Our PC Network boots of the server.  USers can't play with such files.

So the argument that Unix introduces additional complexity with users
and security doesn't apply.

--
Alan Braggins  armb@setanta.demon.co.uk  abraggins@cix.compulink.co.uk
"Any technology distinguishable from magic is insufficiently advanced"


------------------------------

From: tzs@u.washington.edu (Tim Smith)
Subject: Re: Just Computers advertising practices
Date: 6 Aug 1994 03:26:48 GMT

In article <31s05u$njr@nnrp.ucs.ubc.ca>, Shawn Thomas wrote:
>Heh, yeah, but now that I think about it, it wouldn't be so hard to 
>filter out junk mail using a simple text search... just look for the $ sign.
>
><Grin>.

You mean like the $ sign your news software puts in the Message-ID of your
postings?  Better make the filter a little bit more sophisticated...

--Tim smith

------------------------------

From: tjrc1@cus.cam.ac.uk (Tim Cutts)
Crossposted-To: comp.sys.sun.apps,comp.windows.x,comp.windows.x.apps
Subject: Re: WABI vs. SoftWindows?
Date: 3 Aug 1994 18:29:06 GMT

bsa@kf8nh.wariat.org (Brandon S. Allbery) writes:

>In article <31ntke$5ap@lyra.csx.cam.ac.uk>, tjrc1@cus.cam.ac.uk (Tim Cutts) says:
>+---------------
>| robertsw@agcs.com (Wallace Roberts) writes:
>| >so tell me, how well does tdb handle SIGSEGV?  what's that?  i didn't
>| >hear you.
>| 
>| Sorry, I don't even know what SIGSEGV is.  I'm a Windows programmer...
>| I told you.  Ask me about messages, Window procedures and resources
>| and I might be able to answer... :-)
>+------------->8

>You might know it as an "Exception 13".  Or, much more likely, as a crash that
>takes out the whole system including the debugger...  (Well, not under OS/2.
>Usually.)

Oh, you mean the famous GPF.  TDW handles it very well.  It drops me
straight into a disassembled view of the instruction where the GPF
occured, and I can work backwards to try to determine why.  Good
enough for you?

Tim.
-- 
===============================================================================
194, Vinery Rd, Cambridge, CB1 3DS, UK (+44) 223 572622
http://mole.bio.cam.ac.uk/~tjrc1/
===============================================================================

------------------------------

From: patrickr@cs.kun.nl (Patrick Reijnen)
Subject: Re: how to use term 2.04?
Date: Thu, 4 Aug 1994 07:53:50 GMT

In <Ctzqnx.AA4@watdragon.uwaterloo.ca> szhan@watdragon.uwaterloo.ca (Philip Siming Zhan) writes:

>This is a faq question. I read the manpage and Term.HOWTO, README, etc..
>But I can not figure out how to use term. I successfully installed
>term 2.04 (I use minicom and seyon and I have a 2400 modem). i do 
>know how to dial to my remote machine via term. Should I use minicom
>or seyon to connect to the remote? I do not use kermit often bcos it
>is very slow on transfering files. Can someone teach me step-by-step
>how to use term?

>BTW, where to get ftp, telnet, gopher, mail, etc., (clients for term)?
>from `n1, n2' disks?

Hi again,

Look for these clients at sunsite.unc.edu. Directories: /pub/Linux/apps/comm/termstuff and /pub/Linux/apps/comm/term/term/extra.

>Phil

Patrick Reijnen


--
************************* Patrick Reijnen *************************
* Department of Computer Science, Catholic University of Nijmegen * 
* Email:  patrickr@{sci,cs}.kun.nl                                *
* WWW:    http://atlas.cs.kun.nl:4000/homepage.html               *

------------------------------

From: ajayshah@almaak.usc.edu (Ajay Shah)
Crossposted-To: alt.cd-rom,comp.sys.sun.misc
Subject: `mkisofs' ?
Date: 6 Aug 1994 00:54:38 -0700

I recently heard about the existence of `mkisofs', or free Unix
software which creates ISO 9660 file systems which can be put on
CD-R.

Is this for real?

Does it work with "all" CD-ROM recorders or some specific models?
Specifically, is the cheap Ricoh CD-R drive supported?

Does it work with all Unix?  The platforms I care for are SunOS and
Linux.  I can do smalltime porting from one Unix to another.

        -ans.
-- 
Ajay Shah, (213)749-8133, ajayshah@rcf.usc.edu

------------------------------

From: cmay@titan.ucs.umass.edu (Christopher M. May)
Subject: Re: VL-Bus SCSI adaptors
Date: 5 Aug 1994 19:01:59 GMT

Richard Smith (rsmith@vpsoft.demon.co.uk) wrote:
: I'm planning to install Linux shortly but need to know what VL-Bus SCSI
: adaptors are compatible and in particular, Adaptec 1542 compatible controllers
: such as the DC880 SCSI caching VL card.  Also, what about the Adaptec 2842
: VL bus controller although I've heard that it's incompatible with Linux.

: Basically, what is the most suitable VL Bus SCSI controller with decent
: performance?

I've heard good things about the Buslogic card, though my personal
experience is with the Ultrastor 34f.  I believe the Buslogic 455(s?)
emulates an Adaptec 1542 (22?), and I have heard a number of success
stories regarding this controller.  The Ultrastor was a
nightmare to setup under SLS 1.03, as the bootdisk would hang for me
with the controller installed.  The solution was to install onto IDE,
and recompile a suitable kernel after correcting the mailbox port address
to 340h, and reinstall the controller, bingo, off we go.

But, since I sold my U34F, and bought an Adaptec 1742, I haven't looked back.
The setup was much easier.  Run EISA configuration utility, set the 
card to "Enhanced Mode" and the slackware 1.2 kernel booted like a charm,
and recognized all my drives and my CD-ROM right off the bat.  No jumpers,
no kernel recompile.

If you're gonna spend some money on a SCSI controller right now, 
why not buy the twin-port 2742 EISA controller which is out?  Although
the 2nd port will not be supported for a while, I feel confident that it
will be soon, as a driver is currently under development.

VESA := Patch Fix until PCI settles down and becomes supported
EISA := True bus solution, highly researched and developed, tried and true.

Just my $.02
--

-Chris May, Computer Science, University of MA, Amherst
-       Technical Assistant, P.C. Maintenance Lab


------------------------------

From: nevman@server.Berkeley.EDU (Nevin Cheung)
Crossposted-To: comp.os.linux.help
Subject: linux-1.1.39 doesn't wanna compile w/TMC850 Driver
Date: 6 Aug 1994 07:27:07 GMT

I have successfully compiled everything from 0.99.15 to 1.1.37...  Now, 
when I tried to compile 1.1.39, the following errors came up when I tried 
to compile the SCSI drivers...  What's the problem?  Where is "STAT" 
supposed to be?  Thanx...

gcc -D__KERNEL__ -I/usr/src/linux/include -Wall -Wstrict-prototypes -O2 
-fomit-frame-pointer -pipe -m486 -DARBITRATE -DSLOW_HANDSHAKE -DFAST32 -c 
seagate.c 
seagate.c: In function `borken_wait':
seagate.c:263: warning: empty body in an if-statement
seagate.c: In function `internal_command':
seagate.c:675: `STAT' undeclared (first use this function)
seagate.c:675: (Each undeclared identifier is reported only once
seagate.c:675: for each function it appears in.)
seagate.c:675: warning: suggest parentheses around && within ||
seagate.c: At top level:
seagate.c:97: warning: `abort_confirm' defined but not used
make[2]: *** [seagate.o] Error 1
make[2]: Leaving directory `/usr/src/linux/drivers/scsi'
make[1]: *** [driversubdirs] Error 1
make[1]: Leaving directory `/usr/src/linux/drivers'
make: *** [linuxsubdirs] Error 1

-- 
______________________________________________________________________________
Nevin Cheung                                        nevman@server.berkeley.edu
Open Computing Facility General Manager/Staff Member   nevman@ocf.berkeley.edu
Computer Science Undergraduate Association Member     nevman@csua.berkeley.edu

------------------------------

From: mhavlice@myhost.subdomain.domain ()
Subject: dialin for X
Date: 6 Aug 1994 08:08:16 GMT

Hi everyone!
I am interested in setting my box up to allow single users to dialin and
be able to use X applications. So really  I suppose what I am interested in
is being able to have dialin that uses TCP/IP. Any suggestions?

===============================================================================
       mike havlicek                                        _  ,/| 
mhavlice@magnus.acs.ohio-state.edu                         '\`o.O' 
havlicek@ohstpy.mps.ohio-state.edu                          =(_ _)= 
                                                              |U|

                                                             Ack!
===============================================================================


------------------------------

From: terry@cs.weber.edu (Terry Lambert)
Crossposted-To: comp.os.386bsd.misc
Subject: Re: STREAMS  (was I hope this wont ignite ...)
Date: 6 Aug 1994 07:29:48 GMT

In article <Cu2Ey9.2oM@calcite.rhyolite.com> vjs@calcite.rhyolite.com (Vernon Schryver) writes:
] Unfortunately, all of those put and service functions and the generic
] nature of the stream head and scheduler ensure that STREAMS are never
] as fast as sockets.  I think you can make "page flipping" and "hardware
] checksumming" work with STREAMS (two primary techniques for fast
] networking), but I doubt it is possible to make a "squashed STREAMS
] stack" without doing fatal violence to the fundamental ideas of STREAMS.
] The fastest TCP/IP implementations are based on sockets, not STREAMS,
] and they run 2 to 20 times faster (yes, twenty, as in Gbit/sec).

You can build a "stack compiler" that takes I/O and connection
specifications for multiple stacks and "squashes" them into a single
stack with apparently discrete interfaces.  There is at least one
commercial implementation that does this (I would have to look at my
notes at work to see which one).

The page flipping an HW checksumming are both good points.  Another
technique is to "pre-know" how much you nead to read at the card
level; you can do this with incestuous knowledge on a per-protocol
basis in the drivers; this can nearly triple burst rate (but won't
do anything for propagation delay).

Another "cheat" is to start pushing a packet and shove it all the way
down at a high priority.  This isn't combinatorial with "squashing",
and leads to some cute problems unless a lot of thought is taken
beforehand.

One "trick" that does do "fatal violence to the fundamental ideas of
STREAMS" (I like that phrase) is doubly mapping the buffers, pinning
the pages, and passing the address rather than the data itself.  This
requires pre-preparing the page mapping so the kernel and user space
mapping is the same.  Packet assembly at the stram "tail" must take
this into account, but if done correctly, this will save two copies
and a *lot* of page overhead on a 386 (less so on a 486 or other
rational kernel page protecting architecture).

Another "trick" is to preallocate the buffers to include the protocol
header and thus avoid the assembly entirely (leaving only the copy
to card memory, and only that if that is a considration and the card
doesn't DMA from main memory).  This does violence to the buffer return
and the stream head, and generally doubles the buffer memory consumption
(to be safe).  The user space copyin is done into the real buffer as
a unit instead of into "real" (seperate) mbufs.  This techniques is
not usable simultaneously with the previous one, unless the user space
application has incestuous knowledge of the protocol and can handle
skipping the encapsulation (header) data in dealing with the buffer
contents.

STREAMS can be high performance, but, as you note, at almost the
penalty of not being STREAMS any more except in the technical sense.


                                        Terry Lambert
                                        terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.

------------------------------

From: terry@cs.weber.edu (Terry Lambert)
Crossposted-To: comp.os.386bsd.misc
Subject: Re: Usefulness of BSD/Linux Source Knowledge (was BSD vs. LINUX)
Date: 6 Aug 1994 07:35:35 GMT

In article <Cu2ny3.J9o@boulder.parcplace.com> imp@boulder.parcplace.com (Warner Losh) writes:
] In article <31racs$si2@nyheter.chalmers.se> tl@cd.chalmers.se
] (Torbj|rn Lindgren) writes: 
] >Adding SMP was prrobably rather easy. Just put a big lock on the whole
] >kernel... This means that only one process (or thread) can be in the
] >kernel in the kernel a given time, but if that process can run on any
] >processor you have a SMP (per definition, symmetric means that either
] >processor may run kernel tasks, not that more than one process may be
] >in the kernel).
] 
] What you describe is ASMP.  With ASMP, you must block until the
] "Master" CPU is ready to process your kernel request, where with you
] suggestion, you must wait for the current CPU to finish being in the
] kernel.  It sounds like you have a floating Master CPU.  A SMP kernel,
] btw, means that you can have multiple processes in the kernel
] concurrently and not more than one of them is accessing the same
] criticial structures at the same time.  So one could be servicing a
] serial line interrupt while the other one is blitting stuff to a
] remote X server.

Huh?

SMP doesn't absolutely *require* kernel threading and preemption, it's
just more efficient when done that was.

SMP only means that a single master processer is NOT required to take
all your interrupts for you, not that the kernel be multiply reentrant.

I didn't see a requirement in Torbj|r's posting nthat a single processer
take all kernel entrnces, nor all interrupts -- only that only one be
active at a time.  This is not the typical lack of symmetry of ASMP.


                                        Regards,
                                        Terry Lambert
                                        terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.

------------------------------


** FOR YOUR REFERENCE **

The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:

    Internet: Linux-Misc-Request@NEWS-DIGESTS.MIT.EDU

You can send mail to the entire list (and comp.os.linux.misc) via:

    Internet: Linux-Misc@NEWS-DIGESTS.MIT.EDU

Linux may be obtained via one of these FTP sites:
    nic.funet.fi				pub/OS/Linux
    tsx-11.mit.edu				pub/linux
    sunsite.unc.edu				pub/Linux

End of Linux-Misc Digest
******************************
