On 3/7/2011 4:33 PM, Andrew Deason wrote:
> I've published an I-D defining a couple of types to be used for time in
> future RPCs ("RPC refresh", ubik, and the like):
> 
> <http://datatracker.ietf.org/doc/draft-deason-afs3-type-time/>
> 
> The 100 ns resolution and epoch I think is basically what's been
> discussed before, but the handling of varying time resolution is new.
> From discussions with others, though, I feel some kind of resolution
> representation that is tied to the time variable itself is necessary.
> Comments and such welcomed, of course.

Good work.

The only thing that I would add is some text describing conversion
between AFSAbsTime/AFSRelTime and time_t.  In particular, that time_t ->
AFSAbsTime/AFSRelTime is always 1 second resolution and that
AFSAbsTime/AFSRelTime -> time_t must always round down and never up.

Since time resolution information is being included, there should be
discussion that adding AFSAbsTime (100ns) plus AFSRelTime (1s) should
result in AFSAbsTime (1s).

Inclusion of C source code for various operations wouldn't hurt:

 * AFSAbsTime -> time_t
 * time_t -> AFSAbsTime
 * AFSAbsTime + AFSRelTime
 * AFSAbsTime + AFSAbsTime
 * AFSAbsTime - AFSAbsTime
 * AFSAbsTime <=> AFSAbsTime

Jeffrey Altman


Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to