Dan Sugalski writes:
: At 08:40 AM 5/10/2001 -0700, Larry Wall wrote:
: >Dave Mitchell writes:
: >: Content-MD5: FiIz8m/ma8enU5CTBqhsQw==
: >: X-Mailer: dtmail 1.3.0 @(#)CDE Version 1.4.2 SunOS 5.8 sun4u sparc
: >: X-Spam-Rating: onion.valueclick.com 1.6.2 0/1000/N
: >:
: >:
: >: > Briefly: We want the Perl 6 runtime to be an equivalent of the Microsoft
: >: > CLR, so that if you can somehow get bytecode onto it - from whatever
: >: > language - you can run it. So we've got some bytecode that perl can run.
: >: > Now think about what B::Deparse does.
: >:
: >: I knew the intention was to go the B::Deparse route,
: >
: >It was, but the current intention is to pursue Perl 5 translation (and
: >maybe even a limp-along compatibility mode) via an alternate parser
: >sitting beside the standard Perl 6 parser. Then people can convert
: >piecemeal when they see the benefits of Perl 6 outweighing the risks.
:
: There might still be a B::Deparse solution, or a standalone p5 bytecode->p6
: bytecode converter, at least at the start.
Sure, don't want to discourage the B::Deparse folks. In fact, a healthy
competition between them and the p6 parser folks might make both better. :-)
: Both for example reasons
: (potentially to use as, say, a Java bytecode->p6 bytecode converter) and as
: a way to get bytecode streams to feed into the developing p6 interpreter.
That would be one way to build p6. Another approach is to use the p5
interpreter as scaffolding, with some kind of (not necessarily
lightweight) semantic compatibility translation going back and forth.
Then you can write new bits of the p6 interpreter and slip them in
piecemeal while relying on the p5 interpreter to emulate the bits you
haven't written yet. Eventually you've thrown away essentially all the
scaffolding, and then you're done (hah!). This is essentially how I
managed the p4 to p5 transition.
We might need that semantic cushion layer anyway if we want to intermix
p5 and p6 modules during the transition. Intermixing modules isn't in
our long-term best interest, of course, but as a migration aid
I think people would find it extremely helpful. And any way we can
help people with their migration is going to be beneficial in the
long run. It's likely that translated p6 will run faster than emulated
p5 in any event, so I don't think we'll have a long term problem with
people using mixed systems. But it's a critical mass problem, and until
we get the critical mass of p6 modules out there, people won't switch
at all, unless they have a way to mix and match, I suspect.
Larry
