Coda File System

FUSE usability fr Coda?

From: <>
Date: Wed, 19 May 2010 10:28:27 +0200

As a matter of fact, one of the serious hinders for Coda deployment
is the limited number of the platforms with kernel support for Coda.

For the moment these are:
- Linux with "first class" support for Coda;
- Windows XP with hopefully proper functionality (haven't tested the
  latest version);
- FreeBSD with a shaky status (I could use Coda on FreeBSD 7+ when there
  were no other options but it did not fully work, nor was it stable).

Does current Coda work on NetBSD at all?

At least two important platforms have lost kernel support for Coda during the
last years: MacOSX and Solaris.

In the past Coda was totally dependent on its own kernel part as there
was no alternative. Nowadays there is FUSE which has been implemented
on more platforms than Coda. It is probable that FUSE is going to be
considered a usual part of kernel functionality on many platforms.

As an observation, Red Hat for some obscure reason plainly refuses to
enable Coda kernel module in its Enterprise Linux distribution, but it
ships FUSE. (Coda otherwise works fine oh RHEL when you compile
the kernel module yourself).

Historically Coda client utilities have been using pioctls to talk to Venus,
which goes through the kernel (even though it is actually a "pass through"
operation). With the advent of User Land Only Coda (prototyped for MacOSX)
it has been shown that the kernel participation in pioctls is not necessary.

Unfortunately I do not have enough information about the status and
functionality of FUSE on all platforms. Anyway it looks like it could
be used for Coda purposes on Linux (which would cover almost half of the
supported platforms :)

I would appreciate comments from people more familiar with FUSE limitations
on different platforms and hopefully from Jan about the feasibility
of moving Coda to FUSE.

The hope is to get better support for *BSD and also support for MacOSX
and Solaris. It would possibly lead to less kernel-specific development
needed for Coda in the future?

Note that I am not looking for partial functionality, but in the first
hand for full functionality, including e.g. the possibility to run programs
from files on Coda.

With other words, not considering it to coexist with the current
Coda kernel support (which would increase the necessary development efforts)
but rather to fully replace it.

Looking forward to hear comments from kernel and Coda developers.

Received on 2010-05-19 04:29:31