*BSD News Article 19249


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!constellation!osuunx.ucc.okstate.edu!moe.ksu.ksu.edu!hobbes.physics.uiowa.edu!math.ohio-state.edu!usc!elroy.jpl.nasa.gov!decwrl!usenet.coe.montana.edu!osyjm
From: osyjm@cs.montana.edu (Jaye Mathisen)
Newsgroups: comp.os.386bsd.development
Subject: LKM (terry?) Re: Compressing file system ?
Date: 6 Aug 1993 23:38:35 GMT
Organization: Computer Science, MSU, Bozeman MT, 59717
Lines: 35
Message-ID: <23uq1r$bo@pdq.coe.montana.edu>
References: <CBCtwv.4ws@dragon.dsh.org> <23uo8u$9jh@europa.eng.gtefsd.com>
NNTP-Posting-Host: pdq.coe.montana.edu

In article <23uo8u$9jh@europa.eng.gtefsd.com>,
David C. Niemi <niemidc@oasis.gtefsd.com> wrote:
>In article 4ws@dragon.dsh.org, gary@dragon.dsh.org (Gary D. Duzan) writes:
>>In article <23tr2j$3tt@europa.eng.gtefsd.com>, niemidc@oasis.gtefsd.com (David C. Niemi) writes:
>>   gzip includes a utility that does this today.
>
>This is true.  But gzexe is rather slow and un-robust; I would rather see it
>built into the kernel.

One could argue that additional kernel bloat is exactly the *wrong* way to 
go for this kind of stuff.    Witness nearly every OS out there. (Ultrix
springs readily to mind, bloat, bloat, bloat.)

Seems like it should be possible to stuff this in as a user-level process
somehow.  Maybe a user-level NFS thingie.

My knowledge of Mach is minimal, but I think this is something that can
be done under Mach fairly easily. (not the compression file system 
per se, but attaching and making available "kernel-like" things.).

Maybe Terry will jump in here, I think his LKM stuff would do this
kind of thing fairly well, although it's still "stuffing it in the
kernel", it looks a little more elegant, in that it can be
unloaded if necessary.

One option that I don't recall seeing being mentioned is the use of TCX,
which I ran for some time, and it worked great, except for the  problem
of having an extra process hanging around to compress the exe again,
but for grins and giggles, I ran it on my whole disk, and then untcx'd
the minimal files to get going (and all the scripts), and it worked.  Ended 
up with I think a net gain of about 40%.
-- 
 Jaye Mathisen, COE Systems Manager                (406) 994-4780
 410 Roberts Hall,Dept. of Computer Science
 Montana State University,Bozeman MT 59717	osyjm@cs.montana.edu