*** Spam *** [OpenAFS] Why Arla?

[Jeffrey Hutzelman]

On Sunday, December 04, 2005 09:08:16 AM -0700 Kim Kimball <dhk@ccre.com> 
wrote:

> When AFS was still solely a proprietary product of Transarc/IBM there was
> no open source.
>
> Arla was implementing at least the AFS client when no client source was
> openly available.
>
> I expect there's more going on, and don't know what the differences in
> focus are right now.

Well, as Russ noted, Arla's client architecture is considerably different 
from that of OpenAFS.  Because of this difference and because almost all of 
the code was developed independently, Arla and OpenAFS will tend to have 
different bugs.  They'll also gain new features at different times, though 
neither consistently gains them first.

There tends to be some difference in platform focus, largely based on what 
the primary contributors to each project either prefer to use or are paid 
to support.  Arla tends to provide more effective support for the *BSD 
systems; OpenAFS probably provides better support for some of the 
proprietary systems, including Windows, and supports a larger number of 
them.  Linux support is likely about the same for both packages.


Both packages are implementations of the same set of network protocols, and 
they do interoperate (they have to -- Arla doesn't currently include a 
production-quality server; there was server work going on but I don't 
believe it's been touched in some time, and I don't know if they ship it at 
present).  On UNIX systems they try to support the same system call 
interfaces.  The core developers on the two projects are in pretty close 
communication; they read each others' mailing lists, contribute code, and 
get together once or twice a year to work on significant protocol changes, 
interoperability issues, and so on.

-- Jeffrey T. Hutzelman (N3NHS) <jhutz+@cmu.edu>
   Sr. Research Systems Programmer
   School of Computer Science - Research Computing Facility
   Carnegie Mellon University - Pittsburgh, PA