> That seems possible as long as there's some way to map
> sessions to network nodes
well... sessions are much like ports with the exception that you could have
multiple kernels per host so two sessions with the same id could exist in
the host. but that could be remedied at the POE::Kernel level. if
everytime a session is created, it is guaranteed to be unique within that
host (regardless of kernel) then we can safely say:
http://session@host/event
or
http://session/event
in the case of localhost, since
http://host/event
would not be very meaningful. we could of course also do:
http://host/session?event
-----Original Message-----
From: Rocco Caputo [mailto:rcaputo@xxxxxxxxx]
Sent: Thursday, August 19, 2004 4:18 AM
To: poe@xxxxxxxx
Subject: Re: New PoCo Guidelines
On Wed, Aug 18, 2004 at 11:09:33PM -0700, Erick Calder wrote:
> > poe://session@/event
>
> I like poe://session/event but agree that session@kernel makes more sense
> than kernel@xxxxxxxxxx however I don't think session@ is necessary since
> poe:://kernel/event would make no sense anyway.
I understand there's a desire to use POE for applications distributed
across several processes. Up to now, the method for determining which
process a session exists within has been to specify a Kernel ID in the
event name. For example:
kernel/session/event - fully-qualified sample event name
session/event - for events within the current kernel
event - for events within the current session
I objected to poe://session/event based on that assumption. It seemed
that "session" would need to move between the URI's host and path
parts. That would be a proprietary change to an established standard,
which is just wrong no matter how convenient.
If I understand your proposal correctly, however, you're not
suggesting we break URI semantics at all. You might be suggesting the
elimination of a kernel field entirely. That seems possible as long
as there's some way to map sessions to network nodes. The kernel
field would be implied by the session, and the URI semantics are
preserved.
I withdraw my objection if that's the case.
--
Rocco Caputo - http://poe.perl.org/
|
