Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!yale!eagle.wesleyan.edu!amolitor
From: amolitor@eagle.wesleyan.edu
Newsgroups: comp.unix.internals
Subject: Re: Fundamental defect of the concept of shared libraries
Message-ID: <1991May21.211452.43047@eagle.wesleyan.edu>
Date: 22 May 91 02:14:51 GMT
References: <184@titccy.cc.titech.ac.jp> <1991May16.002617.15386@ladc.bull.com> <197@titccy.cc.titech.ac.jp> <7916@auspex.auspex.com> <215@titccy.cc.titech.ac.jp>
Organization: Wesleyan University
Lines: 22

In article <215@titccy.cc.titech.ac.jp>, mohta@necom830.cc.titech.ac.jp (Masataka Ohta) writes:
> As the architechture can not map them, a possible workaround is to flush
> cache/page-table by software at context switch.
> 
	Umm. You generally have to flush this at context switch time anyway,
when you switch the memory map around. Is the phrase 'by software' meaningful
here? I haven't looked at address translation hardware in some years.

	Before saying that sharable libraries are only possible on slow
hardware, I suggest taking a look at the Vax architecture. I would hardly refer
to a vax 9000 as slow, and I point out that it uses sharable libraries.
Further, it is a trivial exercise to sketch a perfectly reasonable
machine/software configuration in which the use of sharable libraries
saves many hundreds of megabytes, or more, of disk.

	Incidentally, my thanks to Mr. Ohta for providing a little levity and
humor in this newsgroup.

	Andrew

> 
> 						Masataka Ohta