*BSD News Article 98580


Return to BSD News archive

Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.Hawaii.Edu!news.caldera.com!enews.sgi.com!nntprelay.mathworks.com!howland.erols.net!newsxfer.itd.umich.edu!newbabylon.rs.itd.umich.edu!not-for-mail
From: Timothy Watson <tmwatson@bagel.rs.itd.umich.edu>
Newsgroups: comp.unix.pc-clone.32bit,comp.unix.bsd.freebsd.misc,comp.unix.bsd.386bsd.misc,comp.os.ms-windows.nt.advocacy,comp.os.linux.x,comp.os.linux.setup,comp.os.linux.advocacy
Subject: Re: Linux needs more work to be done!
Date: 27 Jun 1997 01:15:40 -0400
Organization: What am I doing?
Lines: 87
Message-ID: <xdoradosk1v.fsf@bagel.rs.itd.umich.edu>
References: <5oqbad$1582@ds2.acs.ucalgary.ca>
NNTP-Posting-Host: bagel.rs.itd.umich.edu
X-Newsreader: Gnus v5.4.37/XEmacs 19.15
Xref: euryale.cc.adfa.oz.au comp.unix.pc-clone.32bit:9696 comp.unix.bsd.freebsd.misc:43611 comp.unix.bsd.386bsd.misc:1288 comp.os.ms-windows.nt.advocacy:64628 comp.os.linux.x:65218 comp.os.linux.setup:118760 comp.os.linux.advocacy:103378

cjtan@acs.ucalgary.ca writes:

> I have just put Linux to the ultimate memory management test: start
> up X and then start up a lot of small simple programs from there
> until the physical memory is filled up, then fill up the swap space
> too.  In my case, I just started a lot of xterm and emacs.  At one
> point, when you can't start any more programs, try to shutdown X.
> It is supposed to shutdown all the programs openned in X and then
> shutdown X, returning to the command line.  But guess what happened:
> the machine freezes up!  My question now is: is this a sign of good
> memory management?  If you say that the processor is the one to be
> blamed, then what about the multitasking capability of Linux?  Is
> there any way of bringing up a quickly that can kill off other
> programs like the NT Task Manager?  I can't seem to find one in
> Linux.

You may be able to use <CTRL>-<ALT>-<F1> or <CTRL>-<ALT>-<F2>,
etc. <CTRL> is important, to drop into a virtual console, and then
kill processes. You may have "killall", in which case you can do
something like killall xterm (to kill all the xterms). Another
possibility, if you just end up with a "blank" screen, is that the
libraries of the X system are 5.2.18 and that of the rest of the OS
are 5.4.X (at least, this is when I experience a problem of an
apparently "frozen" system, but the <CTRL>-<ALT>-<F1>, proved that to
not actually be the case). Unfortunately, there are enough differences
across the libc 5 libraries that it would have made greated sense to
link with libraries qualified by a major and minor "soname," that is,
5.4 for example, so the program would look for a library 5.4.

Unfortunately, one may not be able to launch a new process, I expect
this would be true on NT also. Due to the lack of a "hardware
abstraction layer," on the x86 version of linux, it can potentially be
subject to misbehaviour due to a bad driver. This is a known issue,
but putting something like the GGI project's HAL into the kernel would
require a great investment in resources rewriting XFree86 drivers. As
a workaround for emergencies (mostly due to a flaky video card, under
DOS, I had to reboot), there are programs in the SVGALIB suite that
can store/restore video registers, which might help in some cases when
the "reset" command is just not good enough. I have a set of spare
register settings in my /etc directory.

> This posting is just for the sake of discussion, NOT a flame
> bait!  Personally, I think an OS is just a tool, not a religion or
> anything more than a tool.  So use whatever OS that works for your
> situation.  C. J. Tan

Well, personally, I like the idea of basing an OS around standards
whose descriptions are fairly open, especially to the point where more
that one company may be a able to successfuly independently create an
OS.

Part of the problem with Unix is that people "standardized" and
"extended" on AT & T code, or the BSD variant, on which a lot of
network research was created. and another part of the problem is
reportedly the Microsoft Xenix APIs, which were incompatible but
copied. AIX was written and gained a reputation of being hard to port
to, it seems. However, there now is a "single" specification that one
can port to, though the details of how some low-level things, such as
locking, are treated by the OS by default apparently varies by
OS. However, unless one is porting to old AIX or SCO Openserver,
etc. machines, there should be a define that enables XOPEN API
selection. On Linux, after the library in beta, -D_XOPEN_SOURCE will
enable xopen-style functions, this define is also the trigger in some
certified UNIX. Some developers may not know this and continue trying
to write to the default API. Oh, well, strangely, I had to point out
to someone who was actually doing a major project the existence of the
"ldd" command.

Part of the problem that the Wine implementors complain about,
concerning the Win32 implemention, is that it is not always clear how
the dialog boxes, etc. are supposed to work. The documentated APIs are
thus complex at that level. I think it would be interesting that
others would be able to create an unencumbered-by-ownership by a
single corporation, OS. Particularly interesting would be the concept
of an OS so useful that it could eventually be help as an extendable,
base *Operating System* in common, to be used as a common resource.

A kernel like Linux may form the base for such a system. As a
matter-of-fact, I may not be incorrect that the NCOS that Oracle is
using may be able to do just that, with a twist: a Citrix-like
exportation of the . Interestingly enough, it is based FreeBSD.

-- 
________________________________________________________________________
T    i    m    o    t    h    y              W    a    t    s    o    n
   tmwatson@junkmail.umich.edu (get rid of junkmail!)
  __/| Something there is that doesn't love a wall, that wants it down