Dan Ingalls wrote:
My intent wasn't to convince you by my design, but to get insights, where are the possible problems with my design. And now I've gotten much more, namely a raw draft of a more consequent design by a very experienced man (thanks to Allen).
Go, Stephan!
- Dan
But in which direction (see also my other posts)?
Since this is a very ambitious task (at least for me ;-) , but I like such things) resulting in much work, I don't want to work for the trash can here.
Another question: IMO Allen's design is made in the spirit of having a good jitter available. I fear that for the case of not having one the creation of activation records for each block activation (onto the heap) may be a serious performance problem. OK, they may be recycled by holding the last created ones in some list (like for MethodContexts), but at least the logic costs something.
Since there is no official Jitter release so far (at least for Linux) this could be a problem. In addition Jitter implementations come after a normal port of Squeak for new platforms.
Comments?
Greetings,
Stephan