[OpenAFS-devel] convergence of RxOSD, Extended Call Backs, Byte Range Locking, etc.

[Felix Frank]

Hi,

I want to comment on what I'm sure has been fixed since the 1.4.8 times.
Also I apparently don't understand some of the points you make.

> 2) mandatory locking semantics cannot be implemented

What are these? Access to object storage does not bypass the fileserver
(except for actual data transfer). Regular AFS semantics apply, no?

> 6) the GetOSDlocation RPC conflates a number of operations, and should be split

What are these operations? How will callers benefit from the split?
Please note that the GetOSDlocation RPC has undergone some evolution in
the meantime.

> 7) insufficient information is communicated on the wire to support the
> distributed transaction management necessary to ensure atomicity of
> various operations

It would probably be beneficial to get specific in this area. Of course,
distributed transactions are a proper can of worms that should be opened.

> 9) osd location metadata should be cacheable

It is by now, I believe.

> 10) the wire protocol is insufficient to support any notion of data journalling

I'd be grateful for some elaboration on this point. Note that the
protocol wrt. OSD interaction is strongly based on the T10 OSD standard,
with some extensions to allow proper function.

OSDs are very simple entities by design, but I guess you're referring to
the client side here.

> b) changes to cache coherence, especially DV semantics

I'm rather anxious to see your analysis of this issue. I fear that parts
of it may be outdated already. Have you had a change to take a peek at
the RxOSD Subversion repo?

Thank you for the effort (which I suppose is daunting) you're putting
into this.

Cheers
  - Felix