On Thu, 28 May 2009 08:45:42 -0400, Jason House
<jason.james.ho...@gmail.com> wrote:
I'm really surprised by the lack of design discussion in this thread.
It's amazing how there can be huge bursts of discussion on which keyword
to use (e.g. manifest constants), but then complete silence about major
design decisions like thread safety that defines new transitive states
and a bunch of new keywords. The description even made parallels to the
(previously?) unpopular const architecture.
Maybe people are waiting for Walter to go through all the hard work of
implementing this stuff before complaining that it's crap and
proclaiming Walter should have done in the first place?
This seems really unfair to Walter. Then again, I see no indication of
Walter wanting anything else.
Well, there's been a fair amount of previous related discussion. I've
placed a proposal up on Wiki4D
(http://www.prowiki.org/wiki4d/wiki.cgi?OwnershipTypesInD), though since
it was assembled from a bunch of personal notes on the subject and uses
Walter's old suggestion of using 'scope' instead of Bartosz's 'lent' it's
a bit confusing. I'm planning on re-working it, but other deadlines come
first. There's also been a lot of talk about message
passing/future/promise/task/actor/agent based concurrency, data parallel
models such as bulk synchronous programming (BSP) or GPU programming and
auto-parallelization of pure functions. About the only thing needed from
the type system to implement either of these models is the ability for
uniques/mobiles to do a do-si-do type move (which should supported by ref
unique). And BSP/GPU stuff are way too bleeding edge to support in the
language proper yet.
Honestly, I think people are holding back in part because Bartosz has only
started to reveal a threading scheme and so are waiting for him to
complete it, before proverbially ripping it apart.
