From: <ger...@t-...> - 2003-02-28 18:45:55
|
Christophe Rhodes <cs...@ca...> writes: > What I forsee eventually happening is that all of the EARLY-GFs and > EARLY-FOOs and so on will likewise be created as target foos during > this initial !COLD-INIT-FORMS phase, so that by the time the first > LOAD-DEFCLASS (or whatever, it could be LOAD-DEFGENERIC for all I > know) executes there is all of the necessary stuff there. That's more or less the point where I, being lazy, began to ask myself if a PCL bootstrapped in the cold core wouldn't be good enough :-). > Basically, anywhere you see *BOOT-STATE* in PCL will need frobbing, > to make sure that the *BOOT-STATE* 'complete definition is available > by the time a vanilla CLOS top-level form is executed. That should be what I have when building with #+COLD-PCL. > It is indeed not too bad for a start. Having PCL ready before the > type system looks hard, I agree -- which is a bit of a nuisance, > because we have had to reinvent CLOS badly to have an accurate > SUBTYPEP. Yeah, exactly. Maybe it's possible to build up a preliminary primitive type system that's sufficient to get PCL running, and later replace it with the final implementation in CLOS. Something like that. Let's make it all meta-circular :-)! |