Re: Question regarding tags

[Doug Lee]

On Thu, Feb 02, 2006 at 10:30:53AM -0500, Norman Crisp wrote:
> I wondering if anyone knows if the number of tags in a CVS repository
> affects the performance of the repository.  We build nightly, and generate a
> new tag on all files for that build.  We have been working this way for a
> long time and now have 1000's of generated tags, not to mention branches and
> tags within.
> 
> I have not noticed any hit on performance, but are we going to hit a limit
> at some point where the performance will start to degrade?

For operations not involving tags, I see no reason performance would
be reduced by anything except sheer file size, which surely has an
extremely minimal impact on things compared to other
factors--negligible, I'd wager.  Now, if you want to check out 50
files by a tag, and each file has 1000 tags, I'd imagine the search in
each file for the tag you want could take a little time; but again,
things like applying 50 deltas to pull one file will take a lot more
time than searching the top of the file for a tag.  I assume your
average case involves a whole lot more tags than deltas per file, but
since the tags are stored at the top of an RCS file, and since disk
I/O is more often a bottleneck than CPU time, I expect a lot of tags
could be searched in the time it takes to find just one delta.

So I'd say don't worry about it.  Those more expert than I at RCS file
structure should feel free to correct me of course.

-- 
Doug Lee                 address@hidden        
SSB + BART Group         address@hidden   http://www.bartsite.com
"Innovation is hard to schedule." -- Dan Fylstra