Re: The 0.5 Release of Sapphire Is Available

[email protected] (Graham Barr) Mon, 31 Jul 2000 16:08:34 +0100
Newsgroups perl.bootstrap
Message-ID <[email protected]>
On Mon, Jul 31, 2000 at 10:54:11AM -0400, John Porter wrote:
> Nathan Torkington wrote:
> > Modular decomposition of the problem space doesn't mean that each
> > module has to be reusable in contexts other than Perl.  I think a lot
> > of people are missing this.  I think the Perl's RE engine, for
> > instance, is going to assume Perl so that it knows about scalars,
> > callbacks, and so on.
> 
> I hate to have to disagree with you, Nat, but ...
> 
> What does it mean to "know about" scalars, callbacks, etc.?
> If those things are also implemented in modules with well-defined
> interfaces, the RE engine only has to know about those interfaces.
> There's no reason scalars, callbacks, etc. can't be highly
> decoupled from the other modules which use them.

You missed his second paragraph

  You could design an abstract interface to any language, but that's
  going to slow down the Perl version to make it possible to use
  other versions, and I don't (personally, not as project leader)
  like that tradeoff.  I'd be very surprised if that kind of approach
  was adopted.

Which is basically what you are saying and pointing out the
disadvantages.

Graham.