Re: [MUD-Dev] concerning tokenization, compilation, performance, and other fun stuff.

[Joel Dillon <jo#kaffein,troll.no>]

Nate Cain (nate#cotdom,com) spake thusly:
> This would have many advantages over other methods.  The most notable of
> these would be speed.  Objects would be compiled to native machine code,
> and would run directly on the processor.  Also, these objects could very
> quickly and easily be loaded and unloaded as well.  The objects could be
> written in any language for which there was a compiler available for the
> target OS, and could, potentially, allow objects to have access to
> system/os functions that would make objects able to do some very cool
> things.

  And some very nasty things. You need to be careful about security,
at least if you have an open building model.

> This sort of system could also be implemented very easily, and could
> also very easily be multithreaded.  It's a perfect use for the COM
> model, or ActiveX... if you want to go that route.  :-)

  Ick, say byebye to portability ;)

> Has anyone tried this sort of thing?  Are there any docs/papers out
> there on how such a system might work?  if no-one has done it yet, is
> anyone willing to help out? ;-)

  It's been tried, a little - someone on here called Robert Brady was
working on something similar, check the archives. I'd prefer to have a
mixture myself; interpreted languages for ease of debugging, which can
then be compiled on the fly into dynamic libraries for speed.




_______________________________________________
MUD-Dev maillist  -  MUD-Dev#kanga,nu
http://www.kanga.nu/lists/listinfo/mud-dev