Re: Intermediaries - various cases

Mark Baker <[email protected]>
Newsgroups gmane.comp.web.services.ws-arch
Message-ID <[email protected]>
Hi,

On Fri, Sep 27, 2002 at 09:48:19AM -0400, Heather Kreger wrote:
> I am somewhat naive in this space, but I'll weigh in with what I've heard
> in some discussions lately, including the management task force call:

This is my "bread and butter", so to speak.  I build routers.

> I had always thought of intermediaries as transparent to the application.

Some are, but not all.

> In my simple mind there were
>  2 kinds of intermediaries:
>       - those who modify the messages that flow thru them
>             security type ones
>             gateways
>       - those that do not modify it and pass them on
>             routers
>             store/forward

That's one way to characterize them, sure.  But it doesn't tell the
whole story.  Sort of like those "there's two kinds of people in the
world" lines. 8-)

Off the top of my head, you could also characterize them by;

- whether it's identified as recipient of message or not.  i.e. whether
it's a proxy or a gateway
- which administrative domain it's under, sender or receiver

> I guess the term 'transport' intermediary didn't fit for me because many
> intermediaries do 'transport' agnostic things (log, etc.), however, I can
> see the argument for the term because these intermediaries are invoked in
> the process of 'transporting' a message between a requester and provider.
>
> Intermediaries that applications are aware of (i.e. at the application
> layer)

I think all intermediaries that we're talking about are application
layer intermediaries.  Things like TCP PEPs would be an example of a
transport layer intermediary.

> fit more in the work flow/ choreography scenario in my mind

There's many kinds of choreography.  Some of them fit more with

> I had also thought that the set of intermediaries to be processed in the
> course of getting from requester to provider and back again would be
> defined in the configuration of the requester and/or provider when it was
> deployed...

You mean the route isn't part of the message?  So that would have to be
hub-and-spoke style, with the initial message sender actually sending
multiple messages, one to each intermediary?

That's certainly possible, but I wouldn't call that routing, nor would
I call it a good idea in general.

>  so I"m not sure how the scenario of a requester hands out its
> list of intermediearies WITH its message fits into this...

That's what WS-Routing does.

>  Did the
> infrastructure put the list in (transparent to app) or did the app
> specifify them (workflow?).

You can do it both ways.

When you think about it, an intermediary is a *very* generic concept;
just a node that sends and receives messages.  There are many ways in
which a message could be routed to and/or from this node.  Here are
some of the ways that our routing platform supports;

http://www.idokorro.com/routing.html

MB
-- 
Mark Baker, CTO, Idokorro Mobile (formerly Planetfred)
Ottawa, Ontario, CANADA.               [email protected]
http://www.markbaker.ca        http://www.idokorro.com
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.