thanks for the information, i feel much more enlightened about the
tradeoffs involved with running coda after your and jan's posts.

i would have appreciated to see this point in particular in a prominent
place in the documentation:

On 03/28/03, Ivan Popov <pin_at_math.chalmers.se> wrote: 

> and create a new empty one - in some way like you have to choose between
> slow/always-ok or fast/with-limitations on Coda.

i felt very much like i had done something wrong, when in fact, i had
simply to take a tradeoff into account.

a related question i have:

it seems like linux's fsync is the biggest barrier to enjoying better
performance. i'm not sure if it's practical for coda to rethink how
and when it uses fsync (admittedly, i have yet to look through coda's
source code). is there a recommended linux kernel with a nicer
implementation of fsync that i could try out? i'm currently running
2.4.18, and i'm not averse to upgrading or patching the kernel.
unfortunately, i am tied to linux due to the preferences of the
operations people i work with.

-- 

steve simitzis : /sim' - i - jees/
          pala : saturn5 productions
 www.steve.org : 415.282.9979
  hath the daemon spawn no fire?