Yes, Plowman... it's the "without latency" that's not likely to happen. The problem, of course, is that you can't know how long a note is until it's finished, so any samples that rely on duration info -- stac, portato, etc -- need at least _some_ indication of the duration to make a good selection. I was making the comparison to Synful, which has a 1-second delay. I've used this same approach in my program, except I've made it somewhat flexible -- you can have zero delay (or almost zero, given 15 ms, or so, to buffer the sample), but you lose the ability to auto-select stac, port, etc., or you can wind it out to 4-seconds and get more accurate selections for longer notes. You still get legato, though, with the zero delay setting, since in that case the deciding factor actually happens in "real time" (i.e., the overlap of notes).
Now, the big difference with VSL's software will come if it is, in fact, a sequencer. If it is, then it will be able to "look ahead" -- through the entire score, if necessary -- and make all sorts of brilliant decisions. It still wouldn't have to buffer all those decisions, since it could safely buffer just a few notes ahead. This is the sort of option that little hacks like me just don't have when trying to work with existing sequencers -- we can't look ahead, so we have to introduce the "analysis delay". A necessary evil, I suppose.
At any rate, I'm looking forward to the "1 day" page loading later today. Though the hinting seems to be dminishing... nasty tactic -- even MORE suspense!!
J.