On Thu, 15 Apr 1999, Michael C. Richardson wrote:

> 
>   My feelings on the Y2K problem is that it was easier to change programs
> and APIs and typedef's than it was to change on-disk formats.
>   I didn't ask that we *implement* wider times, I just suggested that we 
> make sure that all times in the file systems had space allocated to the 
> 64 bit times. 
>   My understanding of one 64-bit time_t proposal was that it had 48 bits of
> seconds since 1970, or even seconds since 0 or even since -3780 aka 0 on the
> hebrew calendar, since with 47 bits of seconds we get 4456448 years.
>   The other 16 bits become least significant bits for people who need more
> precision.

We're using that precision now, and I don't think 16 bits is really
convenient. I'd vote for a 44/20 split. Then we keep sub-second resolution
in milliseconds, and we're still good for an ungodly number of years. :-)

Take care,

Bill