On Sunday, September 21, 2003, at 02:35 PM, Avi Bryant wrote:

> One possibility we've been discussing is to have downloading all 
> versions
> be the normal method of operation - that is, the main UI for
> opening/loading/merging etc would be built around a browser for a local
> version storage, and MC would periodically just dump any new versions
> from remote repositories into this.  This browser could then color code
> revisions according to whether they're newer or older than your current
> one, provide views by package/repository/author, etc.
>
> I know Julian's a little bit uneasy about this idea, but it's possible
> that as long as you *can* still cherry-pick individual versions from
> remote repositories, he'll let me get away with it. :)

I've begun to think that what we really want is eager copying of 
verison *info* and lazy copying of versions themselves. Then all the 
clients have a clear picture of what the version tree looks like, but 
you only need to ship around the bulk of the data when you actually 
need it for a load or merge.

Some ideas for separating the version info from complete versions:

1) break the mcz format into two parts - one with just the the metadata 
and one with just the data. mcz would still be used for distribution of 
package releases.

2) As Avi mentioned, keep using mcz for the complete version, but 
redundantly have separate files with just the metadata.

3) Leave repositories as they are and have separate mechanisms for 
sharing just version info. A version stream, maybe? Newsgroups?

Colin