On 8/1/2011 1:13 PM, Russ Allbery wrote:
> As nice as it would be to be able to represent old timestamps in the file
> system, we've never been able to before (at least consistently), and I
> think the simplicity benefits for compatibility with current code bases of
> sticking with the POSIX epoch are substantial.
>
> I don't have an opinion on the granularity.  For me, the benefits of
> matching NFS and the POSIX timestamp granularity is fairly evenly balanced
> against the drawbacks of increasing the size of all of our protocol
> packets.

I have a strong desire to ensure that everything that can be represented
in an NTFS file system can be represented in AFS.  I don't care if the
epoch is the same or if the granularity is better than 100ns or not.

I am concerned about the size of status structures given how frequently
they are sent and because the size of the status structure determines
how many FIDs can be specified in a single Bulk Status request.

It would be nice if we could have some form of compressed time
representation that only sent the required number of bits necessary to
represent the required granularity.

Jeffrey Altman

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to