General protocol stack management

Peter Andersson <>
Wed Feb 20 17:49:13 CET 2002


Hi Vlad,

As you've already been told, Markus Kvisth and I did our master thesis on the
subject of general protocol stack interfacing from Erlang a while back. We put
quite a big effort into theory and did not focus as much on implementation. We
prototyped a subset of the ideas that we presented in the report, mainly to
verify that our proposals were feasible for implementation. I think that if you
find functionality described in our paper that you would want to use
practically, then honestly, you're better off writing the code from scratch
than to go digging into our old prototype (implementation should be quite
straightfwd after reading chapter 3, anyway). This holds especially true if you
want to base your stack interface implementation on OTP (which we didn't back
in '97). Anyway, I still have the code and can dust it off for you if you
insist... eh, I mean like.  ;-)

I hope you find the paper interesting. (I do. I must have read it through a
hundred times! ;-). It has a lot of nice little pictures in it, so you
shouldn't have to be disappointed! Seriously, from what you said in your mail,
I think you'll find some interesting things in there. Do let me know! (unless
you hate it).

Best regards

  /Peter  (at OTP, which does indeed belong to some Ericsson company, like Erik
said)



Vlad Dumitrescu wrote:

> Hi all,
>
> I am trying to devise a general way to handle a protocol stack in a way that
> is protocol independent. The closest thing I found to look at was Megaco,
> which is very cool. However, assumptions about the specific protocols are
> spread a little here and there, and also the protocol layers are fixed
> (since it is only designed to handle one protocol).
>
> I wonder if anyone had felt the need for such an application, where once
> could dynamically add, remove or change some of the handlers for the
> different layers in the stack. My reason for doing this is to be able to
> change the encoding and routing algorithms and/or the transport medium, but
> then I thought "why not make it more general?" -- I'd like to find out if
> such generality can be really needed, or maybe it's better to choose the
> custom solution? Maybe such a general aplication already exists, and then
> I'd have a lot to learn from.
>
> Any thoughts or remarks?
> Best regards,
> Vlad
>
> _________________________________________________________________
> Chatta med vänner online, prova MSN Messenger: http://messenger.msn.se




More information about the erlang-questions mailing list