Thursday August 03, 2006
Perl 6 Design Meeting Notes for 02 August 2006
Larry, Allison, Jesse, Nicholas, and chromatic attended the Perl 6 design
meeting on 02 August 2006. These are the minutes.
- working on PDDs
- finished the namespaces one for now
- there'll probably be more modifications for merging classes and
- will it look like the Perl 6 package, module, class, etc?
- probably similar
- what is a namespace outside of a class?
- just a namespace object
- or a namespace role on a module object
- hope to get Chip implementing the exceptions PDD soon
- need to get on some of the other PDDs too
- data types is next
- haven't wrapped that one up yet
- the secondary runloops problem will run large
- anything that wants to do continuations
- threading and that sort of thing
- Patrick also agreed to take on the maintainership of TGE
- I'm happy about that
- I wrote it as a prototype and was going to hand it off to Patrick
- I just kept working on it longer for a while instead
- that's it for me
- making sure that the Synopses and tests stay in sync
- it's about equal work on both ends
- there's a little bit of Haskell hacking in the middle
- also hacking up Pugs to stay in sync with Parrot and adding in the
- also was at a conference last week
- is there a transcript of your SOTO online anywhere?
- I usually give O'Reilly first shot at that
- not much to report besides #parrotsketch
- did manage to get Parrot::Test::Embedded working as a proof of
- the problem with the secondary runloop bothers me
- coming to the conclusion that callbacks are evil
- if you have any intervening C stack, you can't hope to restart without
reconstructing that stack
- instead of thinking in terms of callbacks, you have to think in terms
of sending a message to the primary runloop
- this is why Dan said you can't do things across the boundary if you
write a PMC in C
- I don't think he was thinking in terms of continuations, just
- could you write up your thoughts, Larry?
- I think this'll go into the external API documentation
- I'll be thinking about that the next couple of days while driving
- other people on the list are thinking about it more
- who else is thinking about this?
- Bob Rogers and Piers Cawley (well, in the past Piers did)
- I'm echoing what Bob Rogers said on the list the other day
- he wasn't thinking about message passing
- you can still do calls and returns
- they just have to happen in the main runloop, not a subsidiary
- you need the right instructions to save and restore your state
- and a way for an instruction to say "Here's what to run next"
- almost like a tail recursion optimization
- probably more based on a message queue
- you don't want to slow down your main run loop more than necessary
- I'll make sure Bob didn't get Warnocked
- heard anything else from Mozilla Foundation?
- nothing since the end of OSCON
- I bet Zak is just getting back into things now
- Larry, how is your student doing with the SoC project?
- I don't have any way to measure it other than seeing the checkins
- he probably needs to concentrate now on productization and testing
- how much code can he translate now?
- a fair bit
- it's an open ended project and you can do a lot of refactoring,
especially in relation to objects
- has he tried it on the output of Pugs' Test.pm?
- it would be nice to round-trip that