Hello Sidney, Wednesday, July 20, 2005, 2:53:53 PM, you wrote:
SM> Justin Mason wrote: >> I dunno -- that seems pretty scary. Also, it doesn't fix the problem >> where you have a meta rule in the change which relies on a predicate >> from a previous change, at all. SM> I meant that instead of attaching rules they attach pathname/revision# SM> pairs, not a single revision number for the whole group. But I agree that SM> seems scary. Also, it makes a mess of the svn repository if it can't be all SM> checked out and used as a whole. SM> Dealing with metarules and modifications to them presents a problem in any SM> case. How do we deal with person X submitting a modification to metarule A SM> and proposed rule A1, while person Y submits a different modification to SM> metarule A and proposed rule A2 while person Z submits proposed rule A3 that SM> relies on the existing version of metarule A? This is another example where the mailing list approach works. You're then just looking at three emails, each of which is mass-checked on its own. The problem is in scale -- works great on 5k-10k corpora ... my 250k corpora can't mass-check more than three rules files in a day. Bob Menschel
