Re: [Monotone-devel] clarifications about bookkeeping dir

[Daniel Carosone]

On Wed, Mar 15, 2006 at 02:38:31AM -0800, Nathaniel Smith wrote:
>      (Which is, incidentally, the main thing blocking 0.26 -- the
>      merger is still not well tested.  I hear the argument that we
>      need to get it out the door, but... darn it, tests _are_
>      important.  There's a principle here...)

Absolutely. The clear and self-evident expression of this principle
was one of the things that first attracted me to monotone.

While we're making ourselves clear, let me emphasise that (despite my
enthusiasm to see the improvements released) I'm advocating
concentrating on efforts in the critical path towards getting it out,
such as tests, doc fixes, etc - and defer other non-core features and
changes for next time.  After all, that's typically only a month or so
later - and some more cautious readers of the 0.26 release notes might
wait until 0.27 anyway, just in case.

Making even seemingly simple changes like are being discussed here
takes time, and introduces more things to test and more cases to
handle and more code for bugs to live in.  If we rename the
bookkeeping dir, the presence of the current name is forbidden in
existing revs, but the new name is not. Directory suturing and pivots
with old vs new names introduce more cases to consider and handle and
test. There's also a bunch of trivial but time-consuming work going
over the doc to reflect the change.  Even these discussions take
significant time, though we expected that. :)

I don't even know if any of these cases is hard or not, or what other
issues might be lurking here, if any -- and I don't think we want to
spend time evaluating that Right Now.  I don't see the pressing need,
and I'm unconvinved about the value of the perceived opportunity of
"we're breaking people's revid's and workspaces already and it's the
last chance to make other changes", compared to the need to get 0.26
finished as is.

I just like the symmetry of making 0.26 be all about rosters and
internals; lets make sure we get that right with all the tests we
need, and then we can make 0.27 all about the UI layer and naming if
we want.

All this is just my opinion, of course, and counts for little, however
I'm slightly surprised or uncomfortable to be expressing it in such
detail.  It's the kind of thing I expected would sound like preaching
to the choir, and my apologies if that's so.

--
Dan.

pgpMYNzJ6OAsR.pgp
Description: PGP signature