[Pvfs2-developers] patch (and comments) for pvfs2-viedist

[Phil Carns]

Sounds good to me too.

Murali Vilayannur wrote:
> Yup. It is read-only
> Definitely, sounds like the best approach to me...
> thanks,
> Murali
> 
> On Tue, 14 Mar 2006, Rob Ross wrote:
> 
> 
>>Definitely only need to encode things that have types that span more
>>than one byte.
>>
>>These are read-only values right? So if we just always return them in
>>LEBF we should be ok, regardless of how the server stores them.
>>
>>Rob
>>
>>Murali Vilayannur wrote:
>>
>>>Hi Phil,
>>>
>>>
>>>>If I understand correctly, we have 2 options:
>>>>
>>>>- add an API hook to get the distribution
>>>>- continue grabbing distribution information as an extended attribute,
>>>>but encode it on disk to make it portable
>>>>
>>>>If we got the latter route, then we should probably encode all metadata
>>>>that we store in keyval pairs for completeness (dist, dfile array,
>>>>dirdata handle?, symlink target?, I can't remember what things are
>>>>stored this way currently).
>>>
>>>Yeah, that is true. I am also not sure what the right thing is here..
>>>Or encode it on the fly to lebf as part of the xattr response and keep
>>>disk formats unchanged?
>>>
>>>BTW, does symlink_target have to be encoded? I was thinking that we need
>>>to encode only those keyvals whose bytes could be interpreted?
>>>Thanks
>>>Murali
>>
>>