Path: utzoo!utgpu!utstat!jarvis.csri.toronto.edu!mailrus!uflorida!haven!adm!smoke!gwyn
From: gwyn@smoke.BRL.MIL (Doug Gwyn )
Newsgroups: comp.unix.questions
Subject: Re: How to determine file being redirected to in C
Message-ID: <9803@smoke.BRL.MIL>
Date: 6 Mar 89 12:52:40 GMT
References: <10@sherpa.UUCP> <169@usource.UUCP> <1111@auspex.UUCP> <7869@chinet.chi.il.us> <1121@auspex.UUCP>
Reply-To: gwyn@brl.arpa (Doug Gwyn (VLD/VMB) <gwyn>)
Organization: Ballistic Research Lab (BRL), APG, MD.
Lines: 19

In article <1121@auspex.UUCP> guy@auspex.UUCP (Guy Harris) writes:
>I don't know about the RFS version; life gets more complicated with RFS,
>since, as I remember, if you mount an RFS file system you also end up
>mounting all the file systems mounted on directories in that file
>system.

As I said, the problem looms..

One practical stopgap solution is to include a system ID in the dev_t
(say, the upper half of the datum).  The system ID needs to be unique
across the network.  Given networks such as the Internet with 32-bit
host addresses, one wonders where there's going to be room in dev_t
to store the host identifier.  For the time being a pool of host
addresses can be maintained and the pool index used locally in the
dev_t to identify the site.

(dev_t,ino_t) is just not big enough.  I think we'd better start
phasing in object handles in place of (dev_t,ino_t) at the application
level, before it's too late.