Re: The external interface for the parser piece

>>>>> "NC" == Nicholas Clark <[email protected]> writes:

NC> But on the other hand I also liked Simon Cozen's argument that it should
NC> be easy to embed somewhere else, in which case why do we want to make
NC> the parser have a dependency on the IO library?

I can't see anyway that the dependency can be avoided:

	BEGIN { .... }

If we are parsing perl, the full language (nano, micro, whatever
flavor) will need to be available to BEGIN blocks. Once we have that,
the implementation and design wins are compelling.

We would be ensuring that the IO abstraction works (eating our own
dogfood?), that a useful set of operations are available. Think of
just the suggestions that have come up already, strings as streams,
lazy stream evaluation, all would be useful to a run of the mill perl
script.

We would have only a single 'location' that needs to be aware of
platform dependencies.

Bootstraping could be accomplished with a thin stub layer over FILE *,
with progressive iterations of PerlIO replacing the stub.

<chaim>
-- 
Chaim Frenkel					     Nonlinear Knowledge, Inc.
[email protected]				               +1-718-236-0183