Re: Coroutines are bastards....
Jan-Paul Bultmann <[email protected]>
| Newsgroups | gmane.comp.lang.io |
|---|---|
| Message-ID | <[email protected]> |
On Jun 12, 2011, at 3:14 PM, Steve Dekorte wrote: > > On 2011-06-07 Tue, at 04:36 AM, Jan-Paul Bultmann wrote: > > I mean I love the concept, but it is full of quirks and hiccups and could really need a ton of polish.^^ > > Resuming the right one is difficult, they are hard to handle, and should not be referenced, because at any moment a coroutine could pop of a child and say goodbye without you noticing. Aaaand when debugging is activated on a Coroutine, it doesn't propagate. :/ > > Yes, this is easy to fix. I haven't felt up to working on it but will get to it. > Yes the debugging aspect seems easy to do, though efficiently is another thing. But I think resolving the children issue requires some decisions that should be made by someone with experience :) > Btw, if you think cooperative threads are difficult, > Sorry if I wasn't clear, coroutines are awesome! But it is questionable that they can arbitrarily produce children, that are not really accessible through the parent. The parent should indicate a child generation as well in my opinion :). > you'll find un cooperative threads absolutely unfathomable. :) > yes :D the normal ones suck ^^ But I think a combination of both would be awesome for Io, when the number of CPU cores increases further: One thread per actor (Object), so actors are really running in parallel, and one coroutine per message dispatch. This model would be similar to monitor locks, but way more elegant :D > > Thanks for your patience, > Btw, if we compare io with some other language, they all took about 10 years until they really took off, or died :/, so we should find a killer app for io :D or it may never soar ;) > Steve > >