It would be really nice if someone verified that disconnected startup
really worked, and that such situations survived temporary
reconnections even with old tokens.  IMHO, unless this all works, coda
can't be relied upon for files that you actually need to read when you
aren't on the net -- and I am in practice refraining from really
relying on it.

  I'm still considering adding some form of encryption (using the user's
  password?) to the saved token so that it could possibly be left around
  in a less secure location (publicly readable in /coda?) or could be
  emailed around.

Just save the uid/codaname mapping in rvm, and cons up a fake token on
restart.  There is no particular need to keep state that lets you talk
to the server, I think.  Or is there?

  Finally, I tend to just suspend my laptop all the time. Right now I have
  only 5 days uptime, but only because I just booted into a new kernel.
  My laptop can be suspended for several days without a problem and will
  even do a suspend to disk if the battery is almost completely drained.
  So in my case disconnected operation does get far more testing compared
  to disconnected startup.

But surely your machine crashes occasionally, or your battery runs
down, etc.  That you are getting away with this is nice, but it isn't
a general solution for all usage patterns.  Before realms, this used
to work.  Now, I think the realms resolution needs to happen before
coda is usable at all.

It would also be cool if the rejected tokens were still usable
locally; it would be a drag to lose the ability to see files just
because one got on the net for 3 seconds once.

I'm actually working on a mobility project where that isn't all that
unlikely, except it was more like 30s every 5 minutes in the demo when
the plane with a router flew overhead.  I'd like to use coda in my
next demo (which is not soon).

-- 
        Greg Troxel <gdt_at_ir.bbn.com>