Not File System Specific? (Not Quite) 297Date: Wed, 12 Dec 90 15:07 EST
From: Jerry Roylance <[email protected]>
Subject: Emacs needs all file servers? (was: AB going down)
To: [email protected]^5
Cc: [email protected], [email protected]Date: Wed, 12 Dec 90 14:16 EST
From: Carl R. Manning <[email protected]>Out of curiosity, is there a good reason why Emacs can’t start
up (e.g., on rice-chex) when any of the file servers are down?
E.g., when AB or WH have been down recently for disk
problems, I couldn’t start up an Emacs on RC, despite the fact
that I had no intention of touching any files on AB or WH.Sun brain damage. Emacs calls getcwd, and getcwd wanders down
the mounted file systems in /etc/mtab. If any of those file systems is
not responding, Emacs waits for the timeout. An out-to-lunch file
system would be common on public machines such as RC. (Booting
RC would fix the problem.)Booting rice-chex would fix the problem. How nice! Hope you aren’t
doing anything else important on the machine.
Not Supporting Multiple Architectures
Unix was designed in a homogeneous world. Unfortunately, maintaining a
heterogeneous world (even with hosts all from the same vendor) requires
amazingly complex mount tables and file system structures, and even so,
some directories (such as /usr/etc) contain a mix of architecture-specific
and architecture-dependent files. Unlike other network file systems (such
as the Andrew File System), NFS makes no provisions for the fact that dif-
ferent kinds of clients might need to “see” different files in the same place
of their file systems. Unlike other operating systems (such as Mach), Unix
makes no provision for stuffing multiple architecture-specific object mod-
ules into a single file.
You can see what sort of problems breed as a result:
(^5) Forwarded to UNIX-HATERS by Steve Robbins.