Status update & proposed direction for branch-specific rules

[Andrew Bennetts]

Ian Clatworthy wrote:
[...]
> file during early versions of 1.6. But the core team couldn't agree:
> Martin, John, Alexander and I supported it while Robert and Jelmer
> didn't. From memory, Aaron abstained but I suspect he agreed with Robert
> in principle. In the end, Robert asked in August for .bzrrules to be
[...]

Btw, just for completeness, I'm with Robert/Jelmer: I think this sort of data
ought to be versioned but not as a regular file in the tree (and I think this
is a desireable facility to have in general, because it's easy to imagine us
adding more features than just ignores and filtering rules in the future).

> 3. Design and implement an out-of-tree-but-versioned solution for
>    rules, ignores and anything else that needs it. This may take a long
>    time to do, but then again it may not, depending on how it hangs
>    together.
> 
> I'm going to have a crack at #3 and if that doesn't fly, argue for #1.
> I've had some thoughts over the weekend that I've written those up as a
> draft spec. See http://bazaar-vcs.org/DraftSpecs/MetafileManagement.
> 
> Thoughts?

I think that spec should explicitly mention that the proposed .bzrmeta
directory gives us the flexibility to change the represenation of that
metadata later (e.g. we can change our mind about the ".bzrmeta" name in
later formats without disrupting old format working trees, we have a cleaner
point notice syntax errors in some metadata that should prevent a commit, if
we are conceptually storing data like ignore lists sematically then smarter
merging may be possible, etc etc).

I haven't let all the details sink in fully yet but so far I like it.

-Andrew.