Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!ut-sally!husc6!cmcl2!brl-adm!brl-smoke!gwyn
From: gwyn@brl-smoke.ARPA (Doug Gwyn )
Newsgroups: comp.unix.wizards
Subject: Re: symbolic links are a botch
Message-ID: <6043@brl-smoke.ARPA>
Date: Tue, 30-Jun-87 13:10:27 EDT
Article-I.D.: brl-smok.6043
Posted: Tue Jun 30 13:10:27 1987
Date-Received: Wed, 1-Jul-87 05:43:21 EDT
References: <7879@brl-adm.ARPA> <2211@bunker.UUCP>
Reply-To: gwyn@brl.arpa (Doug Gwyn (VLD/VMB) <gwyn>)
Organization: Ballistic Research Lab (BRL), APG, MD.
Lines: 21

In article <2259@bunker.UUCP> zink@bunker.UUCP (David Zink) writes:
>'..'-as-last part-of-however-we-got-here can become meaningless.  This cannot
>happen under the current implementation.

Since I don't understand that logic (the current symlink implementation does
not handle .. that way), let me paraphrase the argument.  I think you're
saying that if the underlying linkage structure (hard or soft) is rearranged,
the "current working directory" concept acquires some warts (e.g. one's actual
filesystem location could change with the cwd remaining fixed).  This is true
of ANY implementation; the only difference would be the particular nature of
the warts.  I have more than once found that "cd .." (old style) did not get
me into an existing location.  Fortunately this is a relatively unusual
situation.

The only reason that current implementations get away with keeping just the
current inode (and device) is that that could serve as an abbreviation for
the current path.  With /proc, /n/face, and other oddball types of file system,
it is rapidly becoming necessary to find some other way to record the "current
working directory".  A path prefix is certainly the simplest I can think of.
(As I remarked earlier, it can also be generalized nicely to provide a NEW
facility that UNIX hasn't had before.  It also makes getcwd() a simple task.)