SCTP Discussions

Text archives Help


RE: [SCTP-Discussion] sctp enhancements


Chronological Thread 
  • From: "Jim Wimberley" <rrjames AT nortelnetworks.com>
  • To: discussion AT sctp.de
  • Subject: RE: [SCTP-Discussion] sctp enhancements
  • Date: Sun Feb 25 04:00:01 2001
  • List-id: SCTP Discussions <discussion.sctp.de>

Title: RE: [SCTP-Discussion] sctp enhancements

comments below...

> -----Original Message-----
> From: Michael Tüxen [mailto:Michael.Tuexen AT micmac.franken.de]
> Sent: Friday, February 23, 2001 4:28 PM
> To: discussion
> Subject: Re: [SCTP-Discussion] sctp enhancements
>
> Dear Jim,
>
> thank you very much for your comments!
>
> I think it is a good idea not do to anythink if it will not be used.
>
> It would be very helpful to provide the changes for Solaris support.
> We will publish version 1.0.0 shortly. Would it be possible for you
> to provide pathces against that version?

Sure, no problem.

> Another point we are thinking of for a future release (not 1.0.0) is
> to support the glib event loop. Would this be helpful for you?

Not really, unfortunately, but I think that doing that would be perfectly
reasonable for any example applications that you build.

I would think that you would want to keep the stack proper as
independent of event loop as possible. What I've normally seen
for flexible protocol stacks is that the stack itself does not do
anythting that would interfere with it being incorporated in another
application. The stack wouldn't rely on signals, or rely on having the
event loop. What it would do would be supply some primitives and
set some requirements on what the stack requires from the application
using it. My suggestion of exposing the file descriptors and of having
the application "tick" the stack at regular intervals (usually 20 - 100
times a second) is something that I have seen work pretty well. Sure,
every application has to write some glue to incorporate it, but at least
it is possible to incorporate the stack into existing applications/event
loops without modifying the stack itself.

Now, back to the glib event loop. If the stack provided the generic
primatives, and then you had the "glue" on top to bind it into the glib
event loop, I think that would be the best of both worlds. Almost any
existing application could incorporate the stack, or, if you were
starting from scratch, you could just build upon the code that binds
it into the glib event loop.

Thoughts?

> Happy SCTPing
> Michael
>
> --
> e-mail: Michael.Tuexen AT micmac.franken.de
> _______________________________________________
> discussion mailing list
> discussion AT sctp.de
> http://www.sctp.de/mailman/listinfo/discussion




Archive powered by MHonArc 2.6.19+.

Top of page