Coda File System

Re: New clog

From: Ivan Popov <>
Date: Tue, 27 Apr 2004 12:12:05 +0200
On Sun, Apr 25, 2004 at 09:04:24PM +0200, Michael Tautschnig wrote:
> > The fact that Venus does leak those numbers is very confusing as e.g. "ls"
> > interprets them like uid and makes the system administrator believe
> > that they should/could synchronise Coda internal identities numbers
> > with the client systems ones... Pitiable. It is not possible by definition,
> > not by someone's mistake or bad will.
> Would there be a way to - at least - improve those things: I just thought 
> about the following thing (tell me if that is stupid or not feasible at 
> all...):
> - store the unix uid when running clog
> - set the uid of files to that uid later

 - information about files is stored on some server(s),
   not at all affiliated with the client administrator
   (you can use any client in the world to access any realm)
 - Coda has no notion of a "file owner id", that is: no "identity
   which have special rights about the file". Coda has acls instead.

On the other side, Coda stores the file creator's and last file modifier's
Coda identities (the corresponding internal numbers),
and many people believe one of those would be the "file owner uid" in Unix -
while the semantics is very different.

Right now there is no consistency in presentation of "file owner",
sometimes it is the local process uid, sometimes the Coda identity
number, and the result depends on kernel caching of metadata.
Imho Venus instead should present some faked information all the time,
in a non-misleading way.
At most, it could fake (misleading) information for broken programs's sake -
say after an explicit
  cfs fakefileownerforbrokenprograms
so that no one would think it is a recommendable mode. :)

> Then, at least, the output of `ls` would make some sense... 

It would "look better", but it would be wrong.
You could never rely on it - except when the whole world - i.e. the whole
and very private network - is administered by you.

> I still have another very different question: Will the new clog still allow 
> login using tokens stored in some file for working in disconnected mode? If 

You should not need tokens for diconnected mode. Tokens are of use when
you have servers to talk to - not otherwise.

[right now Venus seems to have issues with some transitions between the four
states - connected/disconnected x authenticated/tokenless,
especially if you have tokens at connection which for some reason happen
to be unusable for reintegration - but anyway, you do not need tokens in
disconnected mode]

Yes, you can store a token in a file for later use.

> Does `hoard` allow adding entire directories to the cache? If so, are newly 
> created files in within these directories cached too?

Yes, if you tell it to do so.

Received on 2004-04-27 06:13:52