Return to BSD News archive
Path: sserve!manuel.anu.edu.au!munnari.oz.au!spool.mu.edu!agate!agate.berkeley.edu!cgd
From: cgd@eden.CS.Berkeley.EDU (Chris G. Demetriou)
Newsgroups: comp.unix.bsd
Subject: Re: Shared libs & development in general.
Date: 9 Dec 92 17:10:10
Organization: Kernel Hackers 'r' Us
Lines: 23
Message-ID: <CGD.92Dec9171010@eden.CS.Berkeley.EDU>
References: <1992Dec1.081943.6184@tfs.com> <1992Dec3.031350.23581@ntuix.ntu.ac.sg>
NNTP-Posting-Host: eden.cs.berkeley.edu
In-reply-to: othman@ntrc25.ntrc.ntu.ac.sg's message of Thu, 3 Dec 1992 03:13:50 GMT
In article <1992Dec3.031350.23581@ntuix.ntu.ac.sg> othman@ntrc25.ntrc.ntu.ac.sg (othman (EEE/Div 4)) writes:
> The Joerg's shared lib looks good because it is configurable. It does
>not touch the kernel. Just like nfs which is just a daemon. Why can't we use
>daemons for pc-cons? What is the performace penalty?
nfs is only partially daemon-based...
for instance, the nfs client-side daemon basically does the following:
ignore certain signals
fork the appropriate number of times,
call async_daemon()
which is a syscall which never returns, and does the client-side NFS
handling...
<chuckle> it's *really* running in the kernel...
Chris
--
Chris G. Demetriou cgd@cs.berkeley.edu
"Sometimes it is better to have twenty million instructions by
Friday than twenty million instructions per second." -- Wes Clark