On Tue, Sep 14, 1999 at 02:05:44AM -0500, Christopher P. Lindsey wrote:
> o How similar to Transarc's AFS is Coda? I imagine that the commands
> are close in terms of functionality, but probably differ in name
> (I come from a strictly AFS background)
Not very similar, they both `originate' from about the same starting
point, but both have had several complete overhauls over the past 12
years.
> o Does Coda interoperate with AFS cells? The general answer appears
> to be no, but can I mount AFS volumes from within a Coda cell (and
> if the authentication is different, can I use something icky like IP
> acls as a hack?)
Not a chance, AFS uses RX for the rpc traffic, Coda uses RCP2/SFTP. Even
if you'd (manage to) replace the transport layer, the whole upper layer
interfaces are completely different. Coda servers have to provide
additional information, like version-vectors, so that clients can detect
replication and disconnected operation related conflicts.
Then there is the fs operation logging, (trickle)reintegration,
resolution, which all do not have an AFS equivalent. Also I believe,
AFS currently handles its cache on a block granularity, while Coda has
file-based granularity, which allow us to make certain consistency
guarantees.
> o We have an existing Kerberos V realm in place. Can Coda authenticate
> to this directly, or do we need to create special per-user Coda
> principals or modify the Kerberos database in any special ways? Can
> anyone direct me to K5 specific documentation?
There is an unexportable patch, which allows a user to get a Coda token
without typing a password if he is Kerberos authenticated.
> o If using Kerberos V authentication, can we use Kerberos keytab files
> to obtain tokens? (this ties into the previous question)
Likely, I haven't tried.
> o How difficult is it to port AFS-aware code to work with Coda? Is
> the API similar, or is it another beast altogether?
It depends on what it is aware of, I never heard of this, probably
because I never asked :)
Jan