[OpenAFS] Proposal: OpenAFS foundation to develop AFS server appliance

Chas Williams (CONTRACTOR) chas@cmf.nrl.navy.mil
Sat, 01 Sep 2012 15:03:43 -0400

In message <50424587.6010006@your-file-system.com>,Jeffrey Altman writes:
>End user organizations have over the last decade asked their storage
>vendors to integrate AFS services into the storage products they
>purchase.  The answer has consistently been 'not interested'.

I suspect the answer that you received was "Can you make a business case
for that?"  That appears to be the polite way of saying "go away".

>The Elders have engaged in discussions with the major operating system
>vendors over the years as well.  Those discussions inevitably broke down
>because AFS3 did not satisfy the needs of a First Class file system.
>(No Ext. Attributes, no alt data streams, no byte range locking, no
>mandatory locking, directory limitations, etc.)

Again, I believe this was just a polite way to say "go away".  While
these limitations do exist, they generally don't impact users on a 
day-to-day basis or there are known workarounds.  Some limitations
are present with any enterprise file system though.

>Given the current state of the protocol, the questions surrounding the
>AFS trademarks, the large investments in NFSv4 and CIFS, and the
>relatively small market of the AFS installed base, there is no interest
>in adding AFS3 protocol support to existing hardware storage products.
>Hardware vendors will only be interested in integrating a product that
>addresses the needs of Stanford University. Of course, that is what Your
>File System, Inc. is building.

Operating system vendors have limited resources just like the OpenAFS
community and there isn't any (or very little) demand for a built-in
AFS client.  After all, from their point of view, if you need that you
could just install OpenAFS right?  So, they are not going to devote time
to support a new feature with essentially no users.  Hardware vendors
are not going add a protocol to their storage box unless there is a
perceived need for it.  Since operating systems don't have AFS clients
there is no need.

This is somewhat like the SDTV -> HDTV "conversion".  Adoption was
initially very slow.  No one was buying HDTV sets since no one was
broadcasting HDTV content.  Why broadcast HDTV content if no one has an
HDTV set?  Why would consumers spend extra money on an HDTV set if there
wasn't any content?  This improved after consumers saw the difference
(and it didnt hurt that the government forced the change to reclaim some
of the broadcast spectrum). 

However, for some people SDTV was good enough and would still good
enough in some cases (for instance, HDTV hasnt improved news coverage).
Similarly, AFS isn't a good fit in all cases and sometimes NFS or CIFS
will be sufficient.  But people need to be shown that in some cases, AFS
will clearly be the winner.  What if you have dozens of client machines
with varying levels of administrative attention?  Do you trust all of
your users?  How important is it to protect your data from exposure?
Do you need to share your data with another group?