Re: Tag reorg

[DJ Delorie <dj at redhat dot com>]

> > Well, *I* prefer using an absolute timestamp (date and time) for the
> > snapshot checkout, because otherwise you may be off by a checkout.
> 
> Using -r DATE means the revision at the start of the specified date, which
> is unique AFAICT. How could you possibly be off by one?

As opposed to not using the timestamp for the checkout, and trying to
compute it later.  Plus I keep forgetting whether a date-only is the
start of the date or the end of the date, and which timezone it
applies to.  Plus commit quanta and possible locking oopses.  Plus DST
overlaps (*which* 1:30am did you mean?).  Just pick a safe time in the
recent past (say, 10 minutes ago) when you know it's not near a
checkout.

But I admit to being excessively paranoid ;-)

> I know. But as I said, my point is that maybe we shouldn't think too
> hard of how to organize billions of tags, before we decide if we
> really need those billions.

I hope Dan Berlin doesn't have to think too hard to answer the
question "would putting all those tags in a subdirectory make svn more
efficient in the long run".  Even if it's not a good policy, I'd still
like to know if it's more efficient.

And even if it's not billions of tags (please don't exaggerate),
grouping a handful of tags in a subdirectory may make it easier to
manage a branch, and having ONE entry in "ls branches/" per project is
(IMHO) three times better than having even three entries per project.

And if it tends to make svn three times more efficient, that adds up
quickly.

And I'd rather satisfy my monkey curiosity than have someone fling poo
at me, thankyouverymuch.