On Sat, Oct 19, 2013 at 8:53 AM, Jonathan S. Shapiro <[email protected]>wrote:
> I have no objection to concurrent revisions, but I don't want to put the > concurrency model on the table. Even if we find a non-standard concurrency > model that we like, shared memory concurrency isn't going away and we need > a scheme that works for that case. > I owe Sandro an apology, in that I reacted too strongly. I've been catching up on the GC literature, but I didn't recognize "concurrent revisions" as one of the collector models. Which is kind of embarrassing, because at one time I knew about the Nettles and O'Toole work on multi-version copying. Long story, not worth repeating. What's going through my head is this: Hybrid RC is emerging as a mechanism that offers a fairly continuous range of design options using the same basic infrastructure. That's *such* a seductive notion that it would be a real shame to jump the gun on it and get caught being full of crap. Because I am excited about this, I'm reacting in an overly-conservative way. At some level I'd rather err on the side of shooting myself down than go out with something half-baked and have somebody else discover an embarrassingly obvious flaw. To me, it's frustrating to feel like I'm so close and be tantalized. As long as it's not carried to extremes, that's a good instinct. But I shouldn't take it out on Sandro. In the morning I'll send out a note about where I am on concurrent collection and how to maintain a fairly continuous range of options on collection technologies. shap
_______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
