"PhiHo Hoang" <phiho.hoang at rogers.com> is claimed by the authorities to have written:


> 	Actually, I think we do not need an image to run the bytecode.
> We need an ObjectMemory to load and 'link' the objects (bytecode). The
> (bytecode) InterpreterPlugin will 'run' them with the help of all
> numbered primitives in the PrimitivesPlugin and other named primitives
> in the regular plugins. Did I miss anything ?
Err, well how about the classes to hold the methoddictionaries to hold the compiledmethods? And maybe a few important objects here and there? Smalltalk f'rinstance?

While one could, technically, generate an image from scratch each time you start up (I did a system that did this once a long time ago and I think maybe LittleSmalltalk still does) it is a pretty pointless exercise in general. After all, an image contains all that stuff already compiled, generated, linked together, what have you. It loads at disk-to-memory speeds, requires (maybe) a run through to swizzle pointers and there you are. Much faster. Unless of course your image includes huge amounts of complex startup related code completely uninteresting to most users... 

tim

-- 
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Every program in development at MIT expands until it can read mail.