Re: timestamp resolution

[Joerg.Schilling@xxxxxxxxxxxxxxxxxxx (Joerg Schilling)]

"James Youngman" <jay@xxxxxx> wrote:


> Suppose a filesystem actually represents timestamps in integer units
> of seven seconds since Jan 8, 1935.   If I touch a file at Fri May  4
> 09:53:26 UTC 2007, the event occurs at 326029315.142857.... units of
> time after the epoch.   This then allows me to calculate that the file
> timestamp will be truncated to 326029315 units, which is Fri May  4
> 09:53:25 UTC 2007.   All well and good.
>
> If all I know is that the resolution is 7 seconds and I do not  know
> the Epoch for the filesystem, how can I predict how a timestamp of Fri
> May  4 09:53:26 UTC 2007 will be rounded with a timestamp resolution
> of 7 seconds?

Well, AFAIK POSIX requires a "resultant" Epoche of Jan 1st 1970 
for time().

Your system could not be POSIX then.

The worst currently known filesystem in POSIX systems seems to be the DOS
filesystem with 2 seconds granularity and this seems to be compatble to
the proposal.

I believe that it would be OK if we allow 2 second granularity and
anything that is a whole-numbered fraction of a second.


Jörg

-- 
 EMail:joerg@xxxxxx (home) Jörg Schilling D-13353 Berlin
       js@xxxxxx                (uni)  
       schilling@xxxxxx     (work) Blog: 
http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily