getting the inventory of a revision, or, Why doesn't the API look like I think it should?

[Jamie Wilkinson]

Following on from my previous mail, why does one ask the Branch for a
Tree based on a Revision, instead of the Revision itself?  (I know,
efficiency :)

My mental model went like this:

 # To examine each of the commits in a branch
 for revision in branch:
   do_stuff_to(revision.inventory)

Now, it took me a while to follow the actual chain of objects, and some
hints from other plugins (thanks gannotate! :), thusly:

 * Get the revisions from the branch

 * Get the commit inventory from the revision

 * Oh, you get an inventory from a Tree!  Get the tree...

 * You get the tree from the branch!  Oh, don't need to get the revision
   now.

 * Examine the inventory from the tree given.

Needless to say, that wasn't obvious from the API.  It also feels dirty to
ask the branch for the tree for a particular revision; knowing about Tree
I'd modify my model above to:

  for rev in branch:
    do_stuff_to(rev.tree.inventory)

So, now that I've found my first assumption about how branches and revisions
fit was wrong given that the rest of what I wanted to do was also at odds
with the API, I'm wondering if anyone can explain why? :)

I mean, bzrlib is maturing, I'm just a single hacker battling against it,
and I don't know the internals.  So either I'm way off base with my
assumptions about how the API should look, or everyone else does :)

Am I on crack, or are my expectations reasonable?