Re: Pushing the `gnus-range-*' functions down into the C layer

[Wojciech Meyer]

> I should say that sometimes C is a necessary evil; but it's an evil
> nonetheless. If you can fix this particular case algorithmically, that
> would be loads better.

Yes.  I would say that you can bootstrap a compiler from minimal
interpreter.  Starting from a very basic primitives (especially Lisp is
well suited for it). You can start from assembly, implement Forth, then
go for Lisp, and then you can actually do anything, However at the end
you probably re-implement half of the language again in gradual rise of
the abstraction barrier.  The key point here is to use C in project
likes Emacs, not only for basic language primitives or things that
cannot be done on the Lisp level but for all the critical functions as
well. Deciding when the function should stay in the core is however
another complex subject. So you trade efficiency for having a clean and
minimal kernel.  So yes, getting rid of C is good but some of stuff
should stay in the core. In this case I think the problem is different,
the functions proposed are very specific, and besides efficiency (which
I think would not increase the performance that significantly, because
of a nature of data structures here, and cost of allocation), there
would be a minimal reason for doing it (however it is mine *PERSONAL*
opinion), and also I believe (but maybe I am wrong), it could be fixed
by replacing the algorithm. Nevertheless, I am curious how much it would
increase the performance.

Mine two (euro)cents,
Wojciech

>
> Just my 2 eurocents,
>
> Andy
> --
> http://wingolog.org/
>
>