RE: [AMG]: Causality: A possible remedy to our one-way, request/r esponse debate.

Hi Glen,

> -----Original Message-----
> From: Glen Daniels [mailto:gdaniels@allaire.com]
> Sent: 24 March 2001 00:17
> Subject: Re: [AMG]: Causality: A possible remedy to our one-way,
> request/r esponse debate.
> 
> 
> Stuart:
> 
> I like this a lot,

Great!

> and I'm glad to see you grok the zen of 
> the monistic (as opposed to monastic :)) one-way operation.

Hey... I have never had a problem with one-way operation... its just there
seemed to be something missing from the one we had. I think/hope I've
managed to put my finger on it.

> Two comments:
> 
> 1) I think your notion is right on, but why not just call it
"correlation"?
> "Causality" doesn't seem as clear to me, and it implies, I think, a bit
more
> about the relationship of the messages than "correlation" does.  (I can't
> think of a particular non-causal example of this type of correlation,
> though, I just like the term better.)

The implication I want to capture is that one message arises as a direct
consequence of another. Correlation suggests to me a much 'looser' notion. I
haven't caught up on the thread on "Correlation (was Transaction ID)" yet -
which may help get a firmer handle on this.

> 2) The "sequence" idea might be something you want to tease out into a
> separate concept - it's certainly useful in the multi-response scenario,
but
> wouldn't it be equally useful for a series of related messages being sent
> from a client to a server, say?

I think that this gets back to whether you are thinking alongs the lines of
"associated with" (ie. correlation) or "as a direct consequence of" (ie.
causality).

The "sequence" idea just seemed like a neat way of being able to represent
(later) request/multi-response at very little cost. Certainly, it's not
essential right now and may benefit from more discussion.

> --Glen

Regards

Stuart

<snip/>

Received on Sunday, 25 March 2001 09:11:43 UTC