tortoise caching and crawling

[Martin Pool]

The Bazaar dirstate working tree format and many APIs are oriented
towards walking the whole tree recursively.  (We are generally trying
to reduce this for better scaling but it is the case now.)  And as I
understand it you're also planning to have a long-running bzr python
process.  So I suspect a separate process that does the recursion may
not win much.

Watching for file changes is definitely an interesting idea though
maybe something for a later release?

-- 
Martin <http://launchpad.net/~mbp/>