627 lines
23 KiB
Plaintext
627 lines
23 KiB
Plaintext
From: Digestifier <Linux-Development-Request@senator-bedfellow.mit.edu>
|
|
To: Linux-Development@senator-bedfellow.mit.edu
|
|
Reply-To: Linux-Development@senator-bedfellow.mit.edu
|
|
Date: Sun, 9 Oct 94 02:13:07 EDT
|
|
Subject: Linux-Development Digest #282
|
|
|
|
Linux-Development Digest #282, Volume #2 Sun, 9 Oct 94 02:13:07 EDT
|
|
|
|
Contents:
|
|
Device driver for serial line/SCSI device (Jan Willamowius)
|
|
Re: 1280x1024, Term, and System Lockup! (Olaf Titz)
|
|
Re: Shared library - HOW? (John S. Kallal)
|
|
Re: Linux For Mac (Tony Jones)
|
|
Re: Telnet & ftp freeze! - AND possible FIX (Thomas E Zerucha)
|
|
Re: Beautifying Linux/Xfree (Tom Wilson)
|
|
Re: 1280x1024, Term, and System Lockup! (Yasuo Ohgaki)
|
|
Re: Orchid Soundwave32 (Michael Hipp)
|
|
Re: Flame on the attitude of Linux towards GCC development (Jeff Kesselman)
|
|
Re: Shared Libs: working toward a permanent solution? (H. Peter Anvin)
|
|
Re: LINUX & VESA vs ISA (H. Peter Anvin)
|
|
Re: Flame on the attitude of Linux towards GCC development (Henry Ware)
|
|
Re: Flame on the attitude of Linux towards GCC development (Stormy Henderson)
|
|
Re: Improving SLIP latency under Linux (Al Longyear)
|
|
|
|
----------------------------------------------------------------------------
|
|
|
|
From: jan@janhh.sh.sub.de (Jan Willamowius)
|
|
Subject: Device driver for serial line/SCSI device
|
|
Date: Thu, 6 Oct 1994 12:43:05 GMT
|
|
|
|
I'd like to write a device driver for a scanner that's connected
|
|
to the serial line or the SCSI port.
|
|
|
|
It should be in the kernel (or a loadable module), so it can register
|
|
a character device and can be accessed bei applications using ioctl.
|
|
But I'd like to use "high level" communication with the serial/SCSI
|
|
port that is unavailable to a kernel driver.
|
|
|
|
What would be the best way to implement such a driver ?
|
|
|
|
- Jan
|
|
|
|
|
|
|
|
--
|
|
Jan Willamowius, Nienredder 6, 22527 Hamburg, Germany
|
|
E-Mail: jan@janhh.sh.sub.de
|
|
(if E-Mail fails, please try: willamow@informatik.uni-hamburg.de)
|
|
|
|
------------------------------
|
|
|
|
From: uknf@rzstud1.rz.uni-karlsruhe.de (Olaf Titz)
|
|
Crossposted-To: comp.os.linux.help
|
|
Subject: Re: 1280x1024, Term, and System Lockup!
|
|
Date: 6 Oct 1994 13:24:04 GMT
|
|
|
|
Joseph Bennett - PCD ~ <jbennett@frx146.intel.com> wrote:
|
|
> I am running Linux on my 486 DX/2 66 PCI system. I have been running
|
|
|
|
How much RAM?
|
|
|
|
> Now, however, I am greedy, and have attempted to alter my Xconfig to run
|
|
> X at 1280x1024 resolution.
|
|
>...
|
|
> This one program, however, causes the whole system to LOCK UP! It draws
|
|
|
|
Does this program use extraordinarliy much memory? If yes, perhaps
|
|
increased RAM usage of the X server and/or the program in question
|
|
(does it make the window bigger?) in higher resolution might cause a
|
|
memory hog.
|
|
|
|
Olaf
|
|
--
|
|
___ olaf@bigred.ka.sub.org - uknf@rz.uni-karlsruhe.de
|
|
__ o <a href="http://rzstud1.rz.uni-karlsruhe.de/~uknf/">click</a>
|
|
__/<_ also: s_titz@ira.uka.de - uknf@dkauni2.bitnet - praetorius@irc
|
|
_)>(_)_________ "now i find that most of the time love's not enough in itself"
|
|
|
|
------------------------------
|
|
|
|
From: kallal@stimpy.eecis.udel.edu (John S. Kallal)
|
|
Subject: Re: Shared library - HOW?
|
|
Date: 9 Oct 1994 00:33:06 GMT
|
|
|
|
In article <jwshin.781607962@nitride.EECS.Berkeley.EDU> jwshin@nitride.EECS.Berkeley.EDU (Jinwoo Shin) writes:
|
|
>Can someone tell me the details of Linux shared library stuffs? Like how to
|
|
>create a shared library with gcc? If there is a faq for this, I couldn't find
|
|
>one.
|
|
|
|
You need to get the file
|
|
ftp://tsx-11.mit.edu/pub/linux/packages/GCC/src/tools-2.16.tar.gz
|
|
|
|
Then make and print out the file 'README.ps' for instructions on how to make
|
|
a DLL shared library. I think that you need to talk to Eric Youngdale about
|
|
address assignments for your shared library. The procedure look a bit complex
|
|
for a brand new library.
|
|
However if you can wait a few months, REAL SOON NOW a elf shared
|
|
library toolset is due out. The elf format shared library fixed some of the
|
|
problems with the current DLL shared library system.
|
|
|
|
John Kallal (kallal@udel.edu)
|
|
|
|
|
|
|
|
------------------------------
|
|
|
|
From: tony@odin.sunquest.com (Tony Jones)
|
|
Subject: Re: Linux For Mac
|
|
Date: 8 Oct 1994 10:33:08 -0700
|
|
|
|
Aaron 'Raz' Wrasman (wrasman@duncan.cs.utk.edu) wrote:
|
|
: Actually could I get some info on Linux for the Mac also? My friend
|
|
: has a Mac and was wondering the same thing, he doesn't keep up with
|
|
: these groups though.
|
|
|
|
It would be nice to have something about the status of the PowerMac port
|
|
posted every so often (other than the 3 lines which never change in the projects
|
|
FAQ).
|
|
From reading the group, it seems lots of people have tried e-mailing the
|
|
developers, and their mail was bit-bucketed. Clearly it's annoying for those
|
|
doing the port to be constantly bonbarded with e-mail.
|
|
Maybe an official status posting every 2 months or so, would keep everyone
|
|
happy ????
|
|
|
|
From the total lack of info, I'm starting to wonder if the port isn't in
|
|
trouble ?
|
|
|
|
tony (still searching for hardware info on the PowerMAC)
|
|
--
|
|
Tony Jones (tony@sunquest.com)
|
|
UNIX Technical Resource Group
|
|
Sunquest Information Systems
|
|
|
|
------------------------------
|
|
|
|
From: zerucha@shell.portal.com (Thomas E Zerucha)
|
|
Crossposted-To: comp.os.linux.help,comp.os.linux.misc
|
|
Subject: Re: Telnet & ftp freeze! - AND possible FIX
|
|
Date: 8 Oct 1994 18:24:58 GMT
|
|
|
|
Please undo my previous patch.
|
|
|
|
I think the following code should fix the problem. It works on my system.
|
|
It corrects a locked socket buffer problem with ppp.c (the one with 2.1.2a).
|
|
(change is at the end of ppp_xmit). (compare 3c505.c).
|
|
|
|
---
|
|
zerucha@shell.portal.com - main email address
|
|
|
|
|
|
*** linhb/drivers/net/ppp.c Fri Sep 16 22:09:22 1994
|
|
--- linux/drivers/net/ppp.c Sat Oct 8 13:40:07 1994
|
|
***************
|
|
*** 1760,1767 ****
|
|
ppp_kick_tty(ppp);
|
|
|
|
done:
|
|
if (skb->free)
|
|
! kfree_skb(skb, FREE_WRITE);
|
|
return 0;
|
|
}
|
|
|
|
--- 1760,1769 ----
|
|
ppp_kick_tty(ppp);
|
|
|
|
done:
|
|
+ #if 0
|
|
if (skb->free)
|
|
! #endif
|
|
! dev_kfree_skb(skb, FREE_WRITE);
|
|
return 0;
|
|
}
|
|
|
|
|
|
------------------------------
|
|
|
|
From: ctwilson@mercury.interpath.net (Tom Wilson)
|
|
Crossposted-To: comp.os.linux.misc
|
|
Subject: Re: Beautifying Linux/Xfree
|
|
Date: 8 Oct 1994 22:28:57 -0400
|
|
|
|
|
|
Ummm, not trying to be picky or anything, but you're crediting me
|
|
with the original posters work. I was trying to say basically the
|
|
same thing you were, except that I was adding that to be VUE-like
|
|
you might need motif...myself, I've used fvwm and xfm to set up a
|
|
fairly handy desktop...
|
|
|
|
In article <BCR.94Oct8001740@k9.via.term.none>,
|
|
Bill C. Riemers <bcr@physics.purdue.edu> wrote:
|
|
:>>>>> "Tom" == Tom Wilson <ctwilson@mercury.interpath.net> writes:
|
|
:
|
|
: Tom> In article <372tg0$1ai@huron.eel.ufl.edu>, Alexandra Griffin
|
|
: Tom> <acg@kzin.cen.ufl.edu> wrote:
|
|
|
|
not mine.....
|
|
:
|
|
: Tom> :3) Another idea from HP-VUE... this environment
|
|
: Tom> features a "console :bar" area at the bottom of the screen,
|
|
: Tom> containing buttons to switch :virtual desktops, invocation
|
|
: Tom> icons for commonly-used apps, small icons
|
|
:
|
|
:
|
|
:It already exists. Its called "GoodStuff" and is part of fvwm. For
|
|
:example, I prefere to put stuff on the side. So I have a left
|
|
:"management" area that contains the following:
|
|
:
|
|
|
|
YUP!
|
|
|
|
{CHOMP}
|
|
:By using the side, istead of the bottom, I still have about 1024x910
|
|
:of my 1152x910 display left. Leaving me ruffly a square screen area
|
|
:to work with.
|
|
:
|
|
: Tom> I've been toying with somthing quite similar using fvwm and
|
|
: Tom> xfm...the functionality is quite similar if you don't mind
|
|
: Tom> using fvwm's virtual desktops.
|
|
:
|
|
:Whats wrong with them. I prefere virtual screens to virtual desktops,
|
|
|
|
NOTHING AT ALL. PLease, don't be so sensitive..I'm not talking
|
|
about your children. Personally, I think they're fine. My remarks
|
|
were aimed at the *ORIGINAL POSTER*.
|
|
|
|
:but normally I use a combination of both. i.e. Completely separate
|
|
:projects go on different desktops, the same project overflows to
|
|
:different virtual screens. Since it is a pain sticking windows
|
|
:switching to another desktop and then unsticking them (the only way
|
|
:I know to move windows between desktops) virtual screens tend to
|
|
:be easier.
|
|
|
|
Sounds fine to me.
|
|
|
|
: Tom> :for system functions (logging out...), and space for a
|
|
: Tom> clock, :calendar, Xload bargraph, & other stuff. The
|
|
: Tom> appearance of the bar is :very professional, with little
|
|
: Tom> beveled insets for each item. I'm
|
|
|
|
Once again, this is the *****original poster,****** *NOT* me.
|
|
|
|
:You can arrange your desktop however you want. I agree this should be
|
|
:much easier to configure. It took me quite awhile to come-up with
|
|
:something I think looks just as professional as as the HP-UX
|
|
:environment. Even longer to improve on it. "vuewm" is you can't
|
|
:load your own background, you have to stick to ugly patterns.
|
|
|
|
You can load whatever you please...you should see the variety of
|
|
wallpapers that my coworkers use.
|
|
|
|
:I much prefere being able to have 'xv" load a random picture from
|
|
:CD every 5 or so minuites, so I'm not constantly looking at the
|
|
:same thing.
|
|
|
|
Probably quite possible from VUE, but you have to remember, I
|
|
was the one that was speaking for what was already there,
|
|
not *for* VUE. The overhead is excessive, it's based on motif,
|
|
which, as we all know is *not free*. Nice, mind you, but I'm
|
|
quite satisfied with what I've put together with X, fvwm, and xfm.
|
|
|
|
:
|
|
:What is really needed is:
|
|
:
|
|
: 1. A Null box. i.e. Something that can be used to mark areas for
|
|
: xload, xbiff, and icons even when they aren't present, but as
|
|
: far as the window manager is conserned don't exist.
|
|
:
|
|
: 2. Auto-resume from last session. i.e. Each time I end-up opening
|
|
: several xterms in one screen, emacs somewhere else, Mosaic, ...
|
|
: if fvwm could remember what I had running when I quit and ask
|
|
: me to restart them again, it would be quite a timesaver.
|
|
|
|
This is the one thing that VUE has that I like, except that it doesn't
|
|
query, you. Does a nice job of remembering what had running if you
|
|
ask it to, though. Now, if its desktops were threaded :-).
|
|
--
|
|
/-----------------------------------------------------------------------\
|
|
| Tom Wilson | "I can't complain, but sometimes |
|
|
| ctwilson@rock.concert.net | I still do." |
|
|
| | -Joe Walsh |
|
|
|
|
------------------------------
|
|
|
|
Crossposted-To: comp.os.linux.help
|
|
From: yohgaki@mercury.cair.du.edu (Yasuo Ohgaki)
|
|
Subject: Re: 1280x1024, Term, and System Lockup!
|
|
Date: Tue, 4 Oct 1994 19:19:30 GMT
|
|
|
|
Joseph Bennett - PCD ~ (jbennett@frx146.intel.com) wrote:
|
|
: Hello.
|
|
|
|
: I am running Linux on my 486 DX/2 66 PCI system. I have been running
|
|
: Term over a 14.4kb modem to dial into work, and was running at 1024x768
|
|
: resolution.
|
|
|
|
: Everything has been working just honky-dory. No problems whatsoever.
|
|
|
|
: Now, however, I am greedy, and have attempted to alter my Xconfig to run
|
|
: X at 1280x1024 resolution.
|
|
|
|
: At first, everything was fine. I grabbed one of the examples from the
|
|
: Sample-Xconfig directory, and my monitor (Nanao F550i) and video card
|
|
: (ATI Mach32 PCI) were cool with it. I dialed into work with Term, no
|
|
: problems, and was able to run all but *1* of my X programs with absolutely
|
|
: no hitches.
|
|
|
|
: This one program, however, causes the whole system to LOCK UP! It draws
|
|
: the main window fine, and gets as far as bringing the second window up,
|
|
: but when it attempts to finish drawing it, the system goes dead. The
|
|
: modem stops sending packets, the disk drive stops, mouse and keyboard
|
|
: are dead. I have to push the computer's RESET button because I am
|
|
: completely dead in the water.
|
|
|
|
: The program is "Vantage Spread Sheet", our VHDL simulator.
|
|
|
|
: I tried a new kernel, same problem. I went back to my older 1024x768
|
|
: resolution, and it all worked fine. Which leads me to believe that
|
|
: I am not setting up the video correctly, and this is causing some
|
|
: really unpredictable behavior.
|
|
|
|
: Anyway, here is my video hardware setup:
|
|
|
|
: Nanao F550i 17" monitor
|
|
: ATI Mach 32 PCI video card (2 Meg RAM)
|
|
|
|
: Xconfig line for 1280x1024 resolution:
|
|
|
|
: "1280x1024" 110 1280 1328 1512 1712 1024 1025 1028 1054
|
|
|
|
|
|
: I had also tried this line, but my monitor apparently didn't like it (too
|
|
: high a refresh rate, I'm assuming):
|
|
|
|
: "1280x1024" 135 1280 1312 1456 1712 1024 1027 1030 1064
|
|
|
|
|
|
This is not what you asked but your monitor supports upto 110Mhz. If
|
|
you use higher than 110Mhz dot clock, it just shutdown automatically
|
|
to protect your monitor. So don't use higher than 110Mhz clock.
|
|
Read README files for X for more info. I think good monitors will
|
|
shutdown automatically if you try to run monitor over spec.
|
|
|
|
BTW, those who don't have what you are doing. You can damage your HW.
|
|
Read readme, howto, etc before play around with XFree.
|
|
|
|
: Any help would be appreciated. I thank everybody who already responded
|
|
: to a post of mine for more Xconfig samples, even though they didn't work.
|
|
: The only other piece of information I can give you is that this is
|
|
: program comes up with the message "unknown X server XFree86" or something
|
|
: like that when attempting to bring the program up.
|
|
|
|
: As an aside, I'm kind of surprised my whole system locked up. I really
|
|
: wouldn't have expected that. I just would have thought Term would die
|
|
: or something.
|
|
|
|
: Also, I'm still unclear on the concept of the three horizontal and vertical
|
|
: timing numbers that appear after the pixel value. Is there some mathematical
|
|
: thing which correlates this to numbers you find in the monitor's Owners Manual?
|
|
: I still can't figure it out.
|
|
|
|
There is a detailed documentation how to calc these numbers. You should
|
|
be able to find somewhere in tsx or sunsite. Not easy to find the best
|
|
value, though. Since it's involves some guess work.
|
|
|
|
--
|
|
Yasuo Ohgaki
|
|
e-mail: yohgaki@phoebe.cair.du.edu
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
------------------------------
|
|
|
|
From: zxmhp01@studserv.zdv.uni-tuebingen.de (Michael Hipp)
|
|
Subject: Re: Orchid Soundwave32
|
|
Date: 8 Oct 94 18:44:24 GMT
|
|
|
|
In <PJG.94Oct7155603@tesla.esl.com> pjg@tesla.esl.com (Paul Gyugyi) writes:
|
|
|
|
>In article <CxBDK2.G6n@cosy.sbg.ac.at> chris@cosy.sbg.ac.at (Christian Linhart) writes:
|
|
|
|
[...]
|
|
|
|
>I tried to get the SW32 recognized as a 16 bit MS Sound System compatible
|
|
>card by the new voxware drivers (to play doom). No luck. I can cat
|
|
>to /dev/audio, but doom's sndserver says I don't have the voxware
|
|
>drivers installed.
|
|
|
|
You can setup your card using the pss-stuff from the voxware driver.
|
|
The pss even can upload your .ld file to the card.
|
|
I have managed to run the MSS (ad1848 codec) _without_ installing
|
|
a soundblaster-driver. (btw: then doom works with 16bit sound)
|
|
|
|
Greetings,
|
|
Mike
|
|
|
|
|
|
|
|
--
|
|
// Michael Hipp ,Talstr. 1, 72072 Tuebingen , Germany // Netmaze, the best //
|
|
// Email: mhipp@student.uni-tuebingen.de // 3D/X11 action-game //
|
|
// x400: c=de;a=d400;p=uni-tuebingen;o=student;s=mhipp // ever written!! //
|
|
|
|
------------------------------
|
|
|
|
Crossposted-To: comp.os.linux.help,comp.os.linux.misc
|
|
From: jeffpk@netcom.com (Jeff Kesselman)
|
|
Subject: Re: Flame on the attitude of Linux towards GCC development
|
|
Date: Sun, 9 Oct 1994 00:48:56 GMT
|
|
|
|
I'm not goping to flame you back, because I have some problems myself
|
|
with the 'version .0 always has bugs' view. (I really don't know enough
|
|
about GNU history to have a very intelligent view on this, just a feelign
|
|
that the only products I've ever seen where this was literraly true were
|
|
all made by Microsoft.)
|
|
|
|
On your subejct of 'come on and help us develop' though, i wanted to make
|
|
you aware of anothre side. Myself, I would love to do some Linux kernel
|
|
hacking, but i don't because I work 40+ hours a week programming for a
|
|
living, which only leaves a small amoutn of time (after the chunkc other
|
|
responsabilities take out) to work on my own projects. Given that, at
|
|
the moment, other projects interest me more. As fasr GCC goes, I'm happy
|
|
I have it, Ill be happy to report bugs, but i don't mhave the time to do
|
|
development on it as well. If that were really necessary, Ild go buy a
|
|
commercial compiler. Luckily, I haven't found that to be the case.
|
|
|
|
Just my nickle's worth,
|
|
(Used to be 2 cents, before Reganomics ;) )
|
|
|
|
Jeff Kesselman
|
|
|
|
------------------------------
|
|
|
|
From: hpa@ahab.eecs.nwu.edu (H. Peter Anvin)
|
|
Subject: Re: Shared Libs: working toward a permanent solution?
|
|
Reply-To: hpa@nwu.edu (H. Peter Anvin)
|
|
Date: Fri, 7 Oct 1994 07:40:11 GMT
|
|
|
|
Followup to: <372jc2$edj@andante.aib.com>
|
|
By author: eric@aib.com (Eric Youngdale)
|
|
In newsgroup: comp.os.linux.development
|
|
>
|
|
> The performance penalty is not due to the library being relocated. It
|
|
> is because we lose one of the machine registers so that it can be used
|
|
> to point to the GOT table. The Intel architecture is painfully short of
|
|
> registers, so losing one gives a noticable performance impact.
|
|
>
|
|
[...]
|
|
>
|
|
> The only way to avoid the performance penalty is to modify the compiler
|
|
> to do something like half-pic. THis way the same pic operands are
|
|
> used so but
|
|
> are not referenced to the %ebx register - then you can free up the %ebx
|
|
> register...
|
|
>
|
|
|
|
Well, it seems to me there should be a way of handling this using a
|
|
segment register, instead of wasting a general-purpose register. Set
|
|
up a local descriptor table, then use the base of a currently unused
|
|
segment register (GS, for example), then do any references with
|
|
respect to GS. This doesn't solve the cases where a pointer needs to
|
|
be passed, of course, but I presume that is not the majority of cases,
|
|
and this would free up EBX.
|
|
|
|
/hpa
|
|
--
|
|
INTERNET: hpa@nwu.edu --- Allah'u'abha ---
|
|
IBM MAIL: I0050052 at IBMMAIL HAM RADIO: N9ITP or SM4TKN
|
|
FIDONET: 1:115/511 or 1:115/512 STORMNET: 181:294/1 or 181:294/101
|
|
This article might have been generated by a buggy newsreader.
|
|
|
|
------------------------------
|
|
|
|
From: hpa@ahab.eecs.nwu.edu (H. Peter Anvin)
|
|
Subject: Re: LINUX & VESA vs ISA
|
|
Reply-To: hpa@nwu.edu (H. Peter Anvin)
|
|
Date: Fri, 7 Oct 1994 07:44:10 GMT
|
|
|
|
Followup to: <36stip$nao@news.halcyon.com>
|
|
By author: darkwind@chinook.halcyon.com (C. Joseph Bridwell)
|
|
In newsgroup: comp.os.linux.development
|
|
>
|
|
> I'd like to know whether people installing LINUX have had more, the same,
|
|
> or less problems depending on whether the PC was VLB or ISA.
|
|
>
|
|
|
|
It should make no difference at all; possibly VLB might have sligtly
|
|
less problems if you have > 16 Mb.
|
|
|
|
/hpa
|
|
|
|
--
|
|
INTERNET: hpa@nwu.edu --- Allah'u'abha ---
|
|
IBM MAIL: I0050052 at IBMMAIL HAM RADIO: N9ITP or SM4TKN
|
|
FIDONET: 1:115/511 or 1:115/512 STORMNET: 181:294/1 or 181:294/101
|
|
Microsoft: The Second Evil IBMpire!
|
|
|
|
------------------------------
|
|
|
|
From: hware@bronze.coil.com (Henry Ware)
|
|
Crossposted-To: comp.os.linux.help,comp.os.linux.misc
|
|
Subject: Re: Flame on the attitude of Linux towards GCC development
|
|
Date: 7 Oct 1994 03:56:54 -0400
|
|
|
|
Are you a moron? Postings belong in at most ***ONE*** linux group.
|
|
Flamebait should be directed to col.misc. You may/may not have valid
|
|
points, I'm supporting the boycott of any crossposted message.
|
|
|
|
With patience,
|
|
Henry
|
|
--
|
|
Will hack Linux for virtual beer.
|
|
|
|
------------------------------
|
|
|
|
From: Stormy@Purple.Madness (Stormy Henderson)
|
|
Crossposted-To: comp.os.linux.help,comp.os.linux.misc
|
|
Subject: Re: Flame on the attitude of Linux towards GCC development
|
|
Date: 7 Oct 1994 07:57:04 GMT
|
|
Reply-To: Stormy@Grand.Mother.Com
|
|
|
|
|
|
What was your point?
|
|
|
|
------------------------------
|
|
|
|
From: longyear@netcom.com (Al Longyear)
|
|
Subject: Re: Improving SLIP latency under Linux
|
|
Date: Sat, 8 Oct 1994 02:58:08 GMT
|
|
|
|
eric@pandora.Las-Vegas.NV.US (Eric J. Schwertfeger) replies to someone:
|
|
|
|
>: Actually, I guess there is one thing you could do. You could set
|
|
>: things up so that if an interactive packet gets queued while a bulk
|
|
>: packet is in the middle of transmission, you immediately interrupt
|
|
>: the bulk packet (by sending an end-of-frame character and relying
|
|
>: on the remote end to discard the incomplete frame) and start the
|
|
>: interactive one instead. Ugly, and I don't recommend it for SLIP
|
|
>: (which has no link error detection). It would improve latency
|
|
>: somewhat.
|
|
|
|
As you indicated, don't try it with SLIP. Putting X window protocol
|
|
over a SLIP link is very dangerous without sending the complete and
|
|
error-free frame.
|
|
|
|
The problem is not that the modems are busy sending the frames. The
|
|
problem seems to come from the fact that the modems themselves have a
|
|
buffer. Once you fill this buffer, the modem will refuse to accept
|
|
additional data until the buffer reaches the low water mark.
|
|
|
|
If you disable the modem buffer then you will probably find that your
|
|
performance will DECREASE since the buffer in the modem smoothes out
|
|
the transmission interrupt latency times. (Yes, it occurs on the
|
|
transmitter as well as the receiver.)
|
|
|
|
The PPP software has defined a "frame abort" sequence. (It is coded as
|
|
ESC-FLAG if you wish to dig up the appropriate RFC documents.) If the
|
|
receiver sees this sequence, the frame is to be considered aborted and
|
|
no further processing (including CRC calculations) is to be performed.
|
|
|
|
Then consider the fact that you may decide to abort the FTP data
|
|
transfer frame just short of its normal completion. This would waste
|
|
the time that you have already spent sending the data to this
|
|
point. (Modem buffers will get in the way in determining just how far
|
|
the modem is in progress for sending the frame.) You then put the
|
|
frame back on the transmit queue for later transmission and send the
|
|
SMTP frame. This gets almost through and then comes a telnet character
|
|
frame. So, on hold goes the SMTP frame and now the telnet goes
|
|
out. The telnet frame completes and you then restart to send the
|
|
SMTP. Oops, now another telnet frame. So back it goes on the queue.
|
|
|
|
This is not a very workable solution. I suppose that you could bump
|
|
the priority by one every time that it goes back on the pending queue
|
|
to avoid starving the frames. However, you need to consider the entire
|
|
system, not just the single message. You need to consider the amount
|
|
of time that you spent calculating the vj header compression, and the
|
|
ability to send all of the data.
|
|
|
|
I doubt that this is a workable solution. However, you are certainly
|
|
free to try . . . . :)
|
|
|
|
If you are the only person using the link and you don't want the
|
|
telnet session to be degraded by ftp, then don't start the ftp
|
|
transfer. (Simple, isn't it?) If you are not the only user then you
|
|
don't have this option. However, consider that the person doing the
|
|
ftp transfer wants the data transferred as fast as possible as
|
|
well. You must consider all of the users, not just one.
|
|
|
|
The best solution is to simply choose the priority when you do the
|
|
transmission and then send it in that order.
|
|
|
|
>How about setting the MTU based on the anticipated amount of interactive
|
|
>traffic, and adjusting it based on the actual amount? Or is the MTU set
|
|
>once for a session?
|
|
|
|
The maximum value is set once for the session. You could reduce it,
|
|
but increasing it is definitely not a good idea.
|
|
|
|
--
|
|
Al Longyear longyear@netcom.com
|
|
|
|
------------------------------
|
|
|
|
|
|
** 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-Development-Request@NEWS-DIGESTS.MIT.EDU
|
|
|
|
You can send mail to the entire list (and comp.os.linux.development) via:
|
|
|
|
Internet: Linux-Development@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-Development Digest
|
|
******************************
|