paul> Oh. I guessed (wrongly) that you promoted both the quotes at
paul> once, and was describing a way to "jump back" and make both
paul> changes in a way that wouldn't screw up the IP.
Sorry I didn't realize you were confused about this. I guess since
you were concentrating on contraction cases, the visible difference of
the two timing schemes wasn't apparent.
paul> Sounds like you've got a more localized state machine than I'd
Yes, it was a happy day when I worked the algorithm to have:
1. Single-character/motion latency (it acts after you type the single
next character or make an insertion point motion).
2. Single-character context (it only needs to look at the character
or structural boundary one before and one after).
There's a lot more code complexity in acting on entire words or
phrases. Spell-checking has some of that complexity because it has no
choice but to operate on words. (My original implementation plan was
to mimic spell-checking, but it's scattered over quite a bit of
countryside.)
paul> envisioned. (It'll be interesting to see how you handle some of
paul> the screw cases with abandoned edits, but that's another story.)
As a general engineering principle, I think local code done right
tends to deal better with odd cases than non-local code (or, if not,
what's OO all about?). So, exercise the cases you have in mind and
let's see what happens.
--
[EMAIL PROTECTED] (WJCarpenter) PGP 0x91865119
38 95 1B 69 C9 C6 3D 25 73 46 32 04 69 D6 ED F3
