Linux-Development-Sys Digest #163, Volume #7 Mon, 6 Sep 99 23:14:10 EDT
Contents:
Re: TAO: the ultimate OS (Peter Samuelson)
Re: Initial Linux System Installation (Matthew)
Re: what's a "jiffy" in /proc? (David T. Blake)
Re: Shutdown Problem (Peter Samuelson)
Re: Can I install Linux on an IBM PS/2 model 95 XP 486? (Matan Ziv-Av)
Re: Linux standards compliance (Peter Samuelson)
Re: Shutdown Problem (Peter Samuelson)
Re: IDE for c++ dev? (Johan Kullstam)
Re: TAO: the ultimate OS (Peter Samuelson)
Re: EROS, persistency, ext3fs? (Peter Samuelson)
Re: Max threads and TCP connections? (David Schwartz)
Re: Max threads and TCP connections? (Peter Samuelson)
Re: Can I compile the kernel using a cc other than gcc? (H. Peter Anvin)
Re: TAO: the ultimate OS (Christopher B. Browne)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: alt.os.linux,comp.os.linux.advocacy,comp.os.misc,comp.unix.advocacy
Subject: Re: TAO: the ultimate OS
Date: 6 Sep 1999 18:58:32 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Peter Samuelson <[EMAIL PROTECTED]>]
> > You keep doing this. Someone raises an objection to something
> > you've said and rather than explaining how to overcome the
> > objection you just wave your hand and say "I believe this objection
> > can be overcome." You seem to like rejecting "false dichotomies",
> > making them false by fiat.
[Vladimir Z. Nuri <[EMAIL PROTECTED]>]
> my "opponents" are making them true by fiat. I am merely pointing out
> unjustified assumptions & dogmas that permeate current OS & software
> design. if you consider anything true that I am making "false by
> fiat" I will look at the evidence and show you where you are wrong<g>
No they are not unjustified. This is the difference between you and
them. Your "opponents" keep bringing up valid reasons to doubt what
you are asserting. At least the reasons look valid. At every turn,
you could take the opportunity of demonstrating why the reasons are
*not* in fact valid, or at least giving some evidence which casts doubt
on the reasons. So far you have not done so. You have waved the
reasons away with "Well, I think it's not necessarily true" or "Well,
whoever thinks that just doesn't have enough imagination."
By "fiat" I am not talking about merely asserting something. I am
talking about rejecting someone else's position, a position for which
he gives evidence, without any counter-evidence but merely with your
own assertions. (And ad-hominem attacks, if "has no imagination" can
be construed that way.)
The instance I was commenting on was where I gave my reasons for why I
don't believe a user interface can be one-size-fits-all. To sum it up,
I said that to make a computer easily usable by computer-illiterates,
it has to adopt the conceptual constraints of natural language:
ambiguities, implications, vague abstractions in place of concrete
abstractions, and other communication inefficiencies. These are
unacceptible to people like me who would rather communicate more
efficiently by learning to "think more like a computer". Your response
did not refute my points, but injected a mere opinion:
> > > I believe an OS that is equally pleasing to the novice as well as
> > > the power programmer is inherently possible. I will continually
> > > reject the false dichotomies suggesting otherwise.
Since you don't seem to understand the difference between arguing using
evidence and examples and arguing with bald opinions, I don't think I
have anything more to argue about.
> note above I did not declare anything "true by fiat". I proposed an
> alternative to existing dogma & conventional wisdom.
You may not be "declaring" things by fiat, but you continue to imply
that your opinions (with only your will to back them up) are so
self-evident that anyone who rejects them is small-minded or openly
hostile. As I said, to me your entire argument style is fallacious and
if that's all you know how to do I don't believe I have anything more
to say.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: Matthew <[EMAIL PROTECTED]>
Subject: Re: Initial Linux System Installation
Date: Tue, 7 Sep 1999 00:58:29 +0100
Hello,
Typically, as ever, I went home after posting the message and worked out
how to do this, however I did it slightly differently from the way
outlined below.
After creating a new partition and mounting it in /mnt/new, I created the
directories /mnt/new/mybin, /mnt/new/compile and /mnt/new/lib-extra.
Into mybin I placed a copy of all the files from /bin and (the important
files, i.e. sed bison, but all would do) from /usr/bin.
Into lib-extra I put a copy of the libraries from /lib.
I created /mnt/new/etc and put in a ld.so.conf file that contained
/lib-extra on one line to tell the loader where to get it's stuff.
I could then do
cd /mnt/new; chroot . /mybin/ldconfig; chroot . /mybin/bash
I might have missed something, but after this I got a new shell with the
root directory in the correct place.
Next bit is the twist.
In the original partition in /mnt/new untar the glibc (and any other)
sources (i.e. binutils, fileutils, textutils, etc) and compile them.
chroot to /mnt/new
Create a directory /mnt and then ln -s / /mnt/new
This allows you to cd to /mnt/new/compile (again...) to the make install.
The programs think that they are finishing off what they had begun, but
they are really installing in a different place.
I only got a problem with the C compiler which I had to do last of all
otherwise It put rubbish directories (containing ' ` [ ( ) " you get the
idea) all over the partition.
It's nice to find out now that there *was* an easier way... Thanks for the
info.
My machine is coming on well now :-).
Regards,
Matthew
On 1 Sep 1999, Klaus Schneider wrote:
>Matthew ([EMAIL PROTECTED]) wrote:
>> Hello,
>
>> I am trying to create a small Linux system from scratch, starting
>> with the C library, and then adding in the compiler. I can then start
>> compiling the other few programs I need. The reason for this is to create
>> a system that is as small and as tidy as possible. Although I am using a
>> very hacked version of RedHat 5.1, I don't like the way most distributions
>> fill my hard drive with stuff I don't want.
>
>Me and a friend are doing just the same at the moment; we too don't like
>that all these distributions are so slow and huge, and sometimes hacked
>versions.
>
>Now, what we did was to compile the newest compiler and binutils to use
>on the distribution system (in your case RedHat). Then, we compiled glibc
>kernel and the most essential programs using that self-compiled compiler
>and binutils. These were copied onto a new, clean partition.
>But however, you cannot give --prefix=/mnt/newdist because some programs
>link the destination of some files into themselves, and if you boot
>your new system, /mnt/newdist/... won't work anymore. Instead, we configured
>with --prefix=/usr (or whatever you need) and compiled. To install,
>you can use
> make install prefix=/mnt/newdist/usr exec_prefix=/mnt/newdist/usr
>This will override the values of the make-variables recursively
>during installation.
>
>If you wish, we probably can give you some documentation of our
>work, and we also could cooperate on the work.
>
>Regards,
>Klaus
>
>
------------------------------
From: [EMAIL PROTECTED] (David T. Blake)
Crossposted-To:
comp.os.linux.development,comp.os.linux.help,comp.os.linux,comp/os.linux.misc
Subject: Re: what's a "jiffy" in /proc?
Date: 6 Sep 1999 17:26:01 GMT
Reply-To: [EMAIL PROTECTED]
Mladen Gogala <[EMAIL PROTECTED]> wrote:
> Jiffy is one tick of the system clock (10 msec).
This is correct on single CPU non-alpha linux systems.
But a jiffy is not really a unit of time - more an
interrupt interval. You get twice as many jiffies per
second on a dual cpu system. Four times as many on
a quad, and >10 times as many on an alpha.
--
Dave Blake
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.misc,comp.os.linux.x
Subject: Re: Shutdown Problem
Date: 6 Sep 1999 19:37:39 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Randall Parker <[EMAIL PROTECTED]>]
> I keep saying that the installs ought to treat certain types of
> directories and files as groups whose locations can be aimed at
> particular partitions _as groups_.
> ie be able to say "Put all configuration files on partition xyz" and
> not have to know what are all the file names that are being moved there.
The FHS (IIRC) defines all global configuration files to go somewhere
in /etc. Things that are per-user obviously go in home directories.
Debian is committed to the FHS and Debian maintainers are responsible
for patching any source that doesn't follow it. Sounds more or less
like what you want.
Note that /etc normally *must* be on the root partition since
/etc/fstab and the runlevel files/directories are in there. There are
tricks you could use to get around this, of course.
(This is the big reason Debian can't properly package qmail -- even in
the "non-free" section -- because the license forbits distributing
binaries that don't follow the original config file layout, which of
course does not follow either the FSSTND or the FHS. Debian instead
has a package "qmail-src" to get around this that gets patched and
built after the fact.)
> I may be mistaken: Aren't there parts of Linux that write temporary
> files in places other than /tmp?
There's /tmp and there's /var/tmp and some packages prefer to use $HOME
as a simple way to avoid temp races, but no software should use
anything else. When I find software that uses another location (like
the 2-D CAD package CCDraft for AIX/HP-UX) I do the symlink thing.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: Matan Ziv-Av <[EMAIL PROTECTED]>
Subject: Re: Can I install Linux on an IBM PS/2 model 95 XP 486?
Date: Mon, 6 Sep 1999 22:29:45 +0200
On Mon, 6 Sep 1999, [iso-8859-1] J=FCrgen v.Hagen wrote:
> Alan Smeltzer schrieb:
> >
> > I have an IBM PS/2 model 95 XP 486.This thing uses an IBM SCSI controll=
er
>=20
> as I've written in another posting, it wasn't difficult to
> do this on my 77.
>=20
> However I have had no luck to use svgalib on the MCA machine.
> It complains about PCI type 1 not being supported.
>=20
> Has anybody managed to get sgvgalib running on an MCA?
This is a problem with the et6000 driver , that was fixed in 1.4.0, but it
should not prevent svgalib from working. Does the PS/2 machine have a vga
compatible video card?
--=20
Matan Ziv-Av [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Linux standards compliance
Date: 6 Sep 1999 19:50:22 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Chris Browne <[EMAIL PROTECTED]>]
> > It's not evident that much actual work has been put into rewriting
> > Linux device drivers to conform with the UDI API
[Nix <$}xinix{[email protected]>]
> AFAIK, none at all has.
>
> I doubt it will happen at all (unless the kernel forks) while the Powers
> that Be in the Linux kernel world are so against it. (Oh dear, *what* a
> pity. ;} )
What *may* happen, though, is that the UDI backers (i.e. the ones with
money) get someone to write UDI backend into the kernel and distribute
*that* as a kernel patch. Not that it would likely go into the
standard kernel but at least the original UDI purposes (serving
hardware vendors who refuse to release source and legacy x86 Unix
vendors who want Linux's level of hardware support) could theoretically
be fulfilled. I don't think that in the end it would help either party
very much, though.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.misc,comp.os.linux.x
Subject: Re: Shutdown Problem
Date: 6 Sep 1999 19:26:14 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[M van Oosterhout <[EMAIL PROTECTED]>]
> Note that journalling doesn't prevent data corruption, only metadata
> corruption. IOW, your filesizes will be correct but the data inside
> the file may be totally wrong.
Correct me if I'm wrong: I believe that with journalling, the only data
corruption you can get is to files being modified or recently having
been modified. Metadata is guaranteed to remain consistent. If you
haven't changed a file in the last few seconds or since an fsync(), it
will *not* get corrupted.
[Randall Parker]
> > (my cleaning lady has inadvertently circumvented my UPS protection
> > twice in the last 3 years btw)
You have an important enough setup to use a UPS but you have staff who
don't know you can't just unplug things? Hmmmm....
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
Subject: Re: IDE for c++ dev?
From: Johan Kullstam <[EMAIL PROTECTED]>
Date: 06 Sep 1999 19:11:49 -0400
Hartmut Rosch <[EMAIL PROTECTED]> writes:
> Johan Kullstam wrote:
> >
> > why do you say this? if you are in X, emacs has menus and a mouse
> > interface. it's been this way for years now. if you can
> > point-n-click your way through, e.g., microsoft notepad, you can
> > point-n-click your way through emacs. there is an extensive menu
> > driven customization facility these days. what do you find
> > unintuitive?
> .......
> > --
> > johan kullstam
>
> Well, it is funny to read this. It is not the point whether you're using
> emacs like an editor or like an IDE. Remember the I stands for
> integrated, that means you have an editor, a class browser, an event
> browser, a debugger and oh yes an executing shell.
emacs has all this. not directly built-in but all are available as
slave modes to emacs.
> And a lot more which eases your daily developing work. All with
> graphical enhancements because pictures or images are much more
> easier to grap.
xemacs.
--
J o h a n K u l l s t a m
[[EMAIL PROTECTED]]
Don't Fear the Penguin!
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: alt.os.linux,comp.os.linux.advocacy,comp.os.misc,comp.unix.advocacy
Subject: Re: TAO: the ultimate OS
Date: 6 Sep 1999 20:11:35 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[[EMAIL PROTECTED]]
> > It isn't about money and never has been to people who contributed
> > to Linux.
[EdToy <[EMAIL PROTECTED]>]
> Is that just like: "you'll get your rewards in the NEXT life"?
Mmmmm, no. Many of us find hacking Linux-related things very rewarding
in this life. Think of it as a hobby we let others benefit from.
> > > Don't come to America and ask people to give up their time
> > > without paying them for it. When it boils down to it, that's all
> > > anyone has.
No, nobody can *demand* that I use my time for something without paying
me for it. However they can *ask* and if it's something I find
worthwhile and fun and which I have time for I work on it. This is how
free software development works.
Or did you think no free software hackers were American capitalists? I
assure you many of us are. I don't hack for idealism but for fun.
> How nice of them to build Red Hat something that they can resell. PT
> Barnum phrase applies?
PT Barnum was right, of course, but this isn't about exploiting. Red
Hat sells value that they add to what we have written. For $79 you get
installation media, manuals and tech support. For $0 you get the
opportunity to download the same software but with no media, no manuals
and no tech support. Take your pick.
The fact that Red Hat is raking it in off the sweat of volunteers like
me does not bother me in the least. *I* am not paying anyone to use
Linux because I don't feel I need the added value I would get. I do
not feel exploited, I feel like Red Hat is a mechanism to allow others
to benefit from what I do for fun anyway. Let 'em profit. No loss to
me, after all.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: EROS, persistency, ext3fs?
Date: 6 Sep 1999 20:34:15 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Guido Fiala <"gfiala"@s.netic nospam.de>]
> Especially the fsck-story
> (http://www.eros-os.org/project/novelty.html) gives me some points.
> Would'nt it be good to make the filesystem persistent in the sense of
> EROS, implementing the something like the mentionend checkpoint
> mechanism? After a powerfail, the fsck would know exactly which
> directory to check and repair and would not have to test the entire
> drive.
If the only thing you want is checkpointing of data and metadata going
to disk so as to make fsck's job easy -- you're talking about
journalling. Journalled filesystems are being developed for Linux (sct
was working on one, don't know where he's at at this point) and are
generally considered a Good Idea, except that they're (inherently?)
somewhat slow for reading, which in many cases represents a lot more
filesystem activity than writing.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: David Schwartz <[EMAIL PROTECTED]>
Subject: Re: Max threads and TCP connections?
Date: Mon, 06 Sep 1999 19:01:56 -0700
Bill LN wrote:
>
> Is it possible to maintain 1 million TCP connections from a single
> Linux box? What would the bandwidth and memory and OS ramifications be
> if I try to hold 1 million TCP connections open at once ?
In my experience, 16,000 seems to be the maximum. I encountered some
nasty memory allocation problems in the Linux kernel above that.
> What are the maximum number of threads I can run on a Linux box and
> process low speed data from 10's or 100's of thousands of TCP
> connections? Would I want to forgo threads altogether and their
> associated context overhead and use kernel code to handle each TCP
> connection via callbacks (as opposed to signaling a thread ?
Umm, more threads do not enable you to do more work. Using threads is
perfectly fine, but you don't need a whole bunch of them.
DS
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Max threads and TCP connections?
Date: 6 Sep 1999 21:01:55 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Bill LN <[EMAIL PROTECTED]>]
> Is it possible to maintain 1 million TCP connections from a single
> Linux box? What would the bandwidth and memory and OS ramifications
> be if I try to hold 1 million TCP connections open at once ?
TCP/IP numbers its ports with a 16-bit number, so there's a limit of
65,536 ports per protocol (TCP or UDP). Each client connection
requires one port, though server connections can share. (Presumably
you're talking server connections so you'd be OK there.) In any case
Linux is not designed to scale up to even close to 1 million
connections or threads.
One other thing to consider is where these connections are coming from.
If they're all on the LAN, you start running into trouble with your
Ethernet switches keeping track of a hundred thousand MAC addresses.
If they're coming in from routers, keep in mind the hard TCP/IP limit
of 65,536 connections between your server and each router. So if it's
all non-local traffic, you need at least 16 routers and a very even
distribution of connections from them.
> What are the maximum number of threads I can run on a Linux box and
> process low speed data from 10's or 100's of thousands of TCP
> connections?
The limits are dynamic I believe but more than a couple thousand
threads or processes will start getting you in trouble. poll(2) is
your friend.
> I'm assuming a very powerful Intel machine (or maybe some other
> arch?) with 100mb network.
Think seriously about losing the `S' in SMP and building a server farm.
When your servers start getting higher-end (especially with Intel),
your returns diminish rapidly. If you need tight communication between
different server threads/processes you can use dedicated point-to-point
links of whatever kind between machines.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (H. Peter Anvin)
Crossposted-To: comp.os.linux.hardware,comp.os.linux.portable
Subject: Re: Can I compile the kernel using a cc other than gcc?
Date: 6 Sep 1999 22:12:16 GMT
Reply-To: [EMAIL PROTECTED] (H. Peter Anvin)
Followup to: <[EMAIL PROTECTED]>
By author: [EMAIL PROTECTED] (Randall Parker)
In newsgroup: comp.os.linux.development.system
>
> In article <7pqvso$k6r$[EMAIL PROTECTED]>, [EMAIL PROTECTED]
> says...
> > Hmmmm, how many of those extensions are used in arch-independent code,
> > I wonder? And I know some GNU extensions (__attribute__ comes to mind)
> > were designed to be easy to dispose of in the preprocessor if needed.
> > Some (like typeof) obviously are not.
>
> IMO, the kernel should be gradually migrated to eliminate non-portable
> extensions.
>
Newsflash: you *can't* write a kernel without non-portable
extensions. Why? Because writing a kernel isn't in the scope of the
C standard (in fact, the C standard doesn't let you do any form of I/O
except by calling the library -- fine, now, how do I write the
library?). There are just too many places where you HAVE to know what
the compiler is going to do with your code (which is nonportable) to
be able to implement a kernel "portably".
-hpa
--
<[EMAIL PROTECTED]> at work, <[EMAIL PROTECTED]> in private!
------------------------------
From: [EMAIL PROTECTED] (Christopher B. Browne)
Crossposted-To: alt.os.linux,comp.os.linux.advocacy,comp.os.misc,comp.unix.advocacy
Subject: Re: TAO: the ultimate OS
Reply-To: [EMAIL PROTECTED]
Date: Tue, 07 Sep 1999 01:47:23 GMT
On 6 Sep 1999 20:11:35 -0500, Peter Samuelson <[EMAIL PROTECTED]>
posted:
> [[EMAIL PROTECTED]]
>> > It isn't about money and never has been to people who contributed
>> > to Linux.
>[EdToy <[EMAIL PROTECTED]>]
>> Is that just like: "you'll get your rewards in the NEXT life"?
>
>Mmmmm, no. Many of us find hacking Linux-related things very rewarding
>in this life. Think of it as a hobby we let others benefit from.
The entertaining-enough result is that if there are enough people
contributing a bit here and there, the fact that the software is infinitely
replicable at minimal cost means that the value starts to add up.
>> > > Don't come to America and ask people to give up their time
>> > > without paying them for it. When it boils down to it, that's all
>> > > anyone has.
>
>No, nobody can *demand* that I use my time for something without paying
>me for it. However they can *ask* and if it's something I find
>worthwhile and fun and which I have time for I work on it. This is how
>free software development works.
>
>Or did you think no free software hackers were American capitalists? I
>assure you many of us are. I don't hack for idealism but for fun.
The *real* point is that if it's worthwhile enough for me to develop
something *for me,* that can be adequate enough value to compensate me
for my time.
This is, for the vast majority of software, the reason why software
gets developed. Look back at the claims about there being "more COBOL
than everything else put together."
Seeing as how there's not vast quantities of web sites out there with
COBOL sources, and there aren't a *lot* of COBOL software packages being
hawked, that provides a pretty strong implication that the COBOL code
out there was written for "bespoke" applications where the value had
to *all* come in the first-generation value to the *sole* user.
>> How nice of them to build Red Hat something that they can resell. PT
>> Barnum phrase applies?
>
>PT Barnum was right, of course, but this isn't about exploiting. Red
>Hat sells value that they add to what we have written. For $79 you get
>installation media, manuals and tech support. For $0 you get the
>opportunity to download the same software but with no media, no manuals
>and no tech support. Take your pick.
>
>The fact that Red Hat is raking it in off the sweat of volunteers like
>me does not bother me in the least. *I* am not paying anyone to use
>Linux because I don't feel I need the added value I would get. I do
>not feel exploited, I feel like Red Hat is a mechanism to allow others
>to benefit from what I do for fun anyway. Let 'em profit. No loss to
>me, after all.
Red Hat has to work pretty hard for the continuing revenues.
They have created software that anybody else can take and sell for $2,
as a bunch of vendors have indeed done.
Furthermore, it is entirely practical for someone else to come along and
say "I'll take RHL, tune it to my purposes, and make a few shekels myself."
To any US resident that feels skeptical about that claim, I suggest
visiting the local Sam's Club, where you will see a derivative work that
owes no royalties to RHAT that is called "Linux-Mandrake."
If people get serious torqued-off at RHAT for whatever reason, it would
be a *remarkably* easy process to move on, between:
- Mandrake
- SuSE
- Caldera
- TurboLinux
- Debian
--
People can be set wondering by loading obscure personal patchable
systems, and sending bug reports. Who would not stop and wonder upon
seeing "Experimental TD80-TAPE 1.17, MegaDeath 2.5..."? The same for
provocatively-named functions and variables in stack traces.
-- from the Symbolics Guidelines for Sending Mail
[EMAIL PROTECTED] <http://www.ntlug.org/~cbbrowne/linux.html>
------------------------------
** 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: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.development.system) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Development-System Digest
******************************
