Re: [Gnu-arch-users] Working out a branching scheme [was: tag --seal --fix]

[Aaron Bentley]

Dustin Sallings wrote:

>     I've got them in the thousands.  They get slower as you add 
> patches.  I've done old CVS imports from trees that have taken many, 
> many hours.  Each patch takes a little bit longer than the previous.  I 
> have a tree I'd like to import that cscvs reports as being 11,324 
> changesets.  I haven't bothered trying to import it yet out of fear.  :)
>
>     I *generally* use arch via webdav, though.  I'm sure if I go to 
> import this thing, I'll do it on a local disk.

Yeah, my archives are usually on a local NFS server.  Works fine so far.

>     It was getting to the point where a commit in such a tree simply 
> took a long time.

Okay, I've noticed this over remote 100k connections.

>     I think part of the problem is just the size of the data set that 
> has to be moved around to figure out what the next patch number is 
> (especially over a network).  It might be easier if the tree weren't so 
> flat.

Hmm.  Assume each listing at 80 bytes, and 10,000 patches, and that's 
800k of directory listings.  Sheesh.  But there's an obvious 
optimization: I have patch-10000; does patch-10001 exist? no? version-0? 
 no? Okay, patch-10001 it is.

>     I'm going to stop now before I start suggesting more directory 
> structure changes.  :)

Oh, like a LATEST-REVISION file?  Yeah, you could be hanged for that. :-)

Aaron