On Wed, Jun 27, 2001 at 02:13:40PM -0400, [EMAIL PROTECTED] wrote:
>
> On 27-Jun-2001 [EMAIL PROTECTED] wrote:
> [...]
> > Let me twiddle some code here.
>
> Here it is http://pied.nu/Perl/objects.perl is a rough draft of how I see
> object layers could be implemented.
i cannot come up with an alternative draft right now because i don't have the
time. i just want to say that to me it still needs a lot of time till we
can start coding ...
we should use the opportunity and develop something really good. at take
all the time we need...
>
> Also, http://pied.nu/Perl/scripting.perl has a been changed around a bit
> so that it can use .. and ... operators.
i would suggest:
-give the script type session some states
-states should return wether they were able to complete their task
-state dependencies:
-it should be possible to group concurrent states, state groups are
states themselves below
-specify which states' completion a state depends on
-specify which locks a state needs
-the script type session calculates a schedule and runs it
-waiting on incoming data not covered, could be implemented using some
special state, that waits and triggers another group
-it should be a loop statement though, as with a loop every algorithm can
be executed
-declaration of the machine:
-give states and locks names, dependencies as hashes (lookup performance)
-no new language needed, just some dependency sets
-nice to have features:
-specify rollback states, so that if one state doesn't finish successfully,
the former's rollbacks are called -> very simple transactions
i think this should work. there are special languages that already deal
with concurrent tasks and dependencies and locks ..., maybe we could use
one of these. there are some that are made for realtime environmentes. POE
isn't realtime of course. but we don't need hard realtime, just fair
scheduling and soft deadlines for timeouts etc.
torvald
