Coda File System

CODA gotchas (was Re: CODA and crash recovery)

From: Andrew Kohlsmith (lists) <aklists_at_mixdown.ca>
Date: Sat, 5 Jul 2008 23:46:29 -0400
On July 3, 2008 06:39:32 am u+codalist-p4pg_at_chalmers.se wrote:
> That's a pity. It looks like I have persuaded you that Coda is
> far too complicated and inconvenient, which it is not.

I haven't found CODA too bad, but there are some gotchas that keep biting me 
and this seems as good a thread as any to hijack.  :-)

> The role of home directory becomes different with global file availability.
> Instead of the variable "$HOME" you can use a _constant_ path to where you
> find your files on Coda. (As a lot of software relies on $HOME, either
> reset HOME or put symlinks in the $HOME directory - and you easily
> access your files which are elsewhere on Coda)

You have to be very careful here.  I have found some KDE applications to be 
particularly nasty about this.  Kopete for sure, and I think KMail as well.  
You can give it a symlink but every now and again they get it in their head 
that a symlink is no good, move/delete the symlink and recreate the entire 
directory path for their configuration/logs/etc.

I dug around in the code a bit, and it appears that it's a kdelibs thing for 
checking/creating the config and data directories.  I haven't looked at it 
much in the past months, though.

> If you really want an experience similar to NFS,
> which means "certain clients only but no extra authentication steps",
> then you may e.g. give the users small local home directories
> with the purpose of storing their Coda passwords in a protected way locally
> and running clog at login automatically.
> The rest of their files may comfortably reside on Coda.
> This is more complicated but it works about as good (or as bad) as NFS :)

I have been toying with the idea of the coda driver emitting some kind of 
message (dbus, udev, etc., nothing seems quite right from kernel->userspace) 
when a token does not exist, so userspace can catch it and provide an 
authentication dialogue for a retry.

I think my biggest sticking point (aside from KDE mentioned above) is that 
when I am disconnected, I really do what what Windows 2000 gave me... a 
real "make available offline" mode.  I know the mechanics of it... before I 
disconnect, cache everything.  However it's the sizing of the log that seems 
that I don't have a grasp on.  Logic dictates that I would want a log as big 
as the entire filesystem in order to access anything in the fs when 
disconnected, no?

-A.
Received on 2008-07-06 00:09:43