FAQ
I'm sort of curious about the discussion (threads on solaris) - Chip has
publicly stated (I believe) that threads will *not* be supported in topaz.
Someone correct me if this statement is wrong, but doesn't this mean that
any work on threading will just make the upgrade path from perl5 to perl6 more
difficult?

Ed

Search Discussions

  • Alan Burlison at Nov 11, 1999 at 9:31 pm

    Ed Peschko wrote:

    I'm sort of curious about the discussion (threads on solaris) - Chip has
    publicly stated (I believe) that threads will *not* be supported in topaz.
    Someone correct me if this statement is wrong, but doesn't this mean that
    any work on threading will just make the upgrade path from perl5 to perl6 more
    difficult?
    Personally I'm far from convinced of the utility/stability/practicality
    of threads in perl5. IMHO its virtually impossible to retrofit threads
    to existing software without either crippling performance or leaving
    gaping holes in the system. I voted strongly against them in the first
    place, but I was overruled the majority decision - which is fair enough.

    The only chance we have of getting a robust and scalable threaded perl
    is if they are put in from the ground floor up, i.e. in Topaz, which if
    you are correct isn't going to happen anyway.

    Alan Burlison
  • Tim Bunce at Nov 11, 1999 at 10:42 pm

    On Thu, Nov 11, 1999 at 09:27:25PM +0000, Alan Burlison wrote:
    Ed Peschko wrote:
    I'm sort of curious about the discussion (threads on solaris) - Chip has
    publicly stated (I believe) that threads will *not* be supported in topaz.
    Someone correct me if this statement is wrong, but doesn't this mean that
    any work on threading will just make the upgrade path from perl5 to perl6 more
    difficult?
    Personally I'm far from convinced of the utility/stability/practicality
    of threads in perl5. IMHO its virtually impossible to retrofit threads
    to existing software without either crippling performance or leaving
    gaping holes in the system. I voted strongly against them in the first
    place, but I was overruled the majority decision - which is fair enough. Ditto.
    The only chance we have of getting a robust and scalable threaded perl
    is if they are put in from the ground floor up, i.e. in Topaz, which if
    you are correct isn't going to happen anyway.
    I think Chip, quite sanely, favors the thread-per-interpreter model.
    And I think that's what perl5 will end up with.

    Tim.
  • Alan Burlison at Nov 11, 1999 at 10:51 pm

    Tim Bunce wrote:

    The only chance we have of getting a robust and scalable threaded perl
    is if they are put in from the ground floor up, i.e. in Topaz, which if
    you are correct isn't going to happen anyway.
    I think Chip, quite sanely, favors the thread-per-interpreter model.
    And I think that's what perl5 will end up with.
    Yes, now you come to mention it I recollect him saying something to that
    effect on p6p. I agree - it seems a far better model, and infinitely
    easier to implement. After all, the less shared state there is, the
    fewer the locking problems there will be :-)

    Alan Burlison
  • John van V. at Nov 16, 1999 at 1:22 pm
    :> > The only chance we have of getting a robust and scalable threaded perl
    :> > is if they are put in from the ground floor up, i.e. in Topaz, which if
    :> > you are correct isn't going to happen anyway.
    :>
    :> I think Chip, quite sanely, favors the thread-per-interpreter model.
    :> And I think that's what perl5 will end up with.

    :Yes, now you come to mention it I recollect him saying something to that
    :effect on p6p. I agree - it seems a far better model, and infinitely
    :easier to implement. After all, the less shared state there is, the
    :fewer the locking problems there will be :-)

    After a decade of running servers I personally dont see any problems w/ forking
    as opposed to threading and many advantages, forks can break w/o bringing
    humanity to a halt.

    But... I have an application where I want to share hoh/hoa/aoh/aoa structures
    between server connections (shamelessly cut+pasted from code Tom C gave me).
    The only way to use IPC is to stringify the structure and re-evaluate it in the
    receiving process.

    Also... Though I personally dont care, doesn't NT require a threading model and
    fork so poorly that excluding multi-threading is breaking some of the
    previously laid out rules?


    =====
    John van Vlaanderen

    #########################################
    # CXN, Inc. Contact: #
    # john@thinman.com, www.thinman.com #
    # 1 917 309 7379 (cell, voice mail) #
    #########################################
    __________________________________________________
    Do You Yahoo!?
    Bid and sell for free at http://auctions.yahoo.com

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupperl6-porters @
categoriesperl
postedNov 11, '99 at 8:53p
activeNov 16, '99 at 1:22p
posts5
users4
websiteperl6.org

People

Translate

site design / logo © 2018 Grokbase