[Vm-dev] Marvin and Cog

[Casey Ransberger]

Marvin seemed to require some extra prims, but I haven't been able to establish why -- if anyone can point me  in the right direction, it would be much appreciated. 

I'm curious: some of the same techniques that make Self performant are present in Cog; what's the remaining tech gap look like? Is it reasonable to think about using Cog to run an object system built around prototypes rather than classes, or is the class assumption baked into the object engine in a way that would require a lot of working around?

I am not a VM guru (obviously) by any stretch of the imagination, and am as such quite wary of making assumptions, but I'd really like to play with bootstrapping a prototype based object model in my nice familiar Squeak environment. Frankly there are some things I'd drop from Self and keep from Squeak. 

The approach I've seen wherein one has a class for every object bugs me somewhat in that it can confuse the hell out of the next programmer, who is likely expecting regular class oriented behavior whilst sitting in front of a class based system. This has been an issue in my work life with Ruby's "eigenclasses" which are pretty much that.

I'd like to experiment with a prototypal model of my own instead, and am looking to scope the challenge. Any advice will be met warmly!