apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@rowe-clan.net>
Subject Re: UUID logic required in APR?
Date Tue, 16 Apr 2002 20:18:20 GMT
At 10:27 AM 4/16/2002, Karl Fogel wrote:
>Greg Stein <gstein@lyra.org> writes:
> > The Windows UUIDs are a helluva better than ours in terms of uniqueness. We
> > compensate, but Windows does a much better job.
>
>Or if we find that really unclean, then abstract one key portion of
>the uuid code and keep it in APR, but put all the uuid interfaces
>(wrappers around that key portion) into apr-util, which can depend on
>apr for platform-specific stuff.
>
>So then apr-util is the official provider of uuid functionality, but
>depends on apr to implement it.

The real question - exactly -what- is the benefit to moving into apr-util?

If we are talking uniquity, there are many machine/os/cpu specific tricks
that could be employed to gain a safely unique ID [across all apr-based
processes] with minimal interlocking pain.  Let's leave sleeping dogs lie
and focus on improving on the shortcomings that Greg just pointed out.

Bill


Mime
View raw message