europython sprint report wrapup

[Michael Hudson]

Aaron Bentley wrote:
> Robert Collins wrote:
>> To wrap up the lithuanian post-europython sprint...
> 
> Thanks for keeping us in the loop.
> 
>> Jelmer has 'branch -r -5.. working, which will only copy the data for
>> the last 5 revisions. This is the key starting point for
>> ShallowBranches/HistoryHorizons. Woot!
> 
> In London, Jelmer and I discussed why we were arguing so strongly about
> History Horizons.  We concluded that Jelmer wants a related feature:
> Lazy Repositories.
> 
> Unlike History Horizons, Lazy Repositories make a branch dependent on
> its upstream branch for old revisions, and do not provide a hard
> guarantee that revisions past the history horizon will *never* be
> stored.  If old revisions are required, (e.g. to perform a merge,) they
> will be fetched and stored in the local repository at that time.

I have to admit that I'd become slightly confused about terminology in
this area, but now I see that, yes, lazy repositories and history
horizons are different things (and both are a better name than "shallow
branches").  Lazy repositories seem more generally useful to me
though... they allow a developer to get a Bazaar branch of (say) CPython
without the restrictions of a lightweight checkout but also without
having to download the entire history in one go, which was what I
understood to be the main use case.

Cheers,
mwh