> Hello folks. We're evaluating the usefullness of the NT/AFS client.
>Our NT guys are looking at it and complaining about the fact that the file
>manipulation interface requires use with the NT 3.51 file manager rather
>than the newer explore. We're also seeing strange crashes when we
>authenticate to an AFS cell and try to use the explorer. Yet the explorer
>works when no tokens are present.
While I can't speak for these problems, I have just discovered that
the NT AFS client _internally_ is a SMB protocol translator. By
that I mean to NT, AFS looks like a SMB server ... which just
happens to be running on the same machine, but with a different
NetBIOS name (groovy, eh?). The AFS client just translates the
SMB requests into the appropriate AFS RX calls. This is instead
of acting like a "real" filesystem to NT (which is what the DFS
client does). This, combined with the memory-based cache, puts
reliability and performance into the proverbial outhouse (it's
about as bad reliability-wise as the AFS-NFS translator). That
might be why you're having problems using Explorer with AFS (but
don't quote me; I'm not an NT expert by any means).
However, I don't think the NT/AFS product is dead by any means.
It's pretty obvious from the copyrights on the AFS pages that they
suffer the same problem as AFS itself; no work has been put into
them in the last few years. At Decorum this year Transarc said
they were renewing AFS development, and working on the problems
with the NT AFS client. So I think things are going to improve ...
but I don't know when.
--Ken
