Re: Proposal for Describing Web Services that Refer to Other Web Services: R085

On Fri, Apr 25, 2003 at 12:07:11PM -0400, Amelia A. Lewis wrote:
> > But for *all* of those systems, it is possible to design a URI scheme
> > (perhaps more than one) such that all the necessary information *is*
> > contained in the URIs (either by value or by reference to a standard,
> > as I mentioned).
> 
> This, though, is where I think we part company.

Good.  I'm glad we got that far. 8-)

>  While it is *possible*
> to design such schemes, it is not always practical.  In particular, when
> other means already exist and are preferred, efforts to promote a
> URI-based syntax tend to stall, stagnate, and fail.
> 
> *Can* be != is.  Also != *should* be, in my opinion.

Perhaps.  All I can state with confidence is that in *every* large scale
system *I* have studied, it has been the case that it was not only
practical, it was easy.

> > I think I also mentioned that every successful large scale distributed
> > system (that I've looked it anyhow, which is many) has this property.
> 
> Depends on what you mean.  If you mean that all successful large scale
> distributed systems use URIs for addressing, I do not agree.

Right.  I didn't mean that.

>  If you
> mean that they all use standardized addressing, it almost goes without
> saying.

What I meant was that not only do they have standardized addressing,
but those addresses comprise sufficient information to interact with
what they identify in the context of that system.  e.g. phone numbers,
http URIs, etc.. (though URIs - well designed ones anyhow - are usable
outside the context of the system that created them).

> This discussion started from a question of whether WSDL addressing
> syntax ought to permit only URI, or ought to permit more complex address
> definition types.  As Mike Champion pointed out, in order to work with
> stuff that currently exists, some of which cannot be identified by URI,
> provision ought to be made to permit more complex types.  This was the
> point in Arthur Ryman's original proposal that I believe you originally
> objected to.  I objected to your objection, on the grounds that while
> URI *might* one day be universal, at present it is not.
> 
> Summary: to support services that are not URI-locatable, it should be
> possible to use more complex addressing syntax, rather than restricting
> address syntax to URI alone.

They can identify any thing, if that's what you mean by "universal".

If you mean that URIs aren't universal in the sense that it isn't
practical to turn any addressing scheme into them, I can't disprove
that, but I can challenge you to provide an example where it would not
be practical.

MB
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca
Web architecture consulting, technical reports, evaluation & analysis

Received on Saturday, 26 April 2003 23:17:43 UTC