--- Steve Jones <[EMAIL PROTECTED]> wrote:

> >  You need to do the same standardization effort for both styles and
> I claim that defining a media type is a lot easier than defining an
> API.
> 
> But that isn't a claim backed up by experience of distributed systems
> which has been successful via APIs but you are saying the new way
> _could_ be easier but don't have the proof point. You might be right,
> but I need data to back that opinion up.

I suppose it's "proof by inspection" but one could also count the
number of internet nodes under different agency and observe how the
netwrok is used.  Specifically, I'd look at the properties of the
popular application protocols that have a variety of implementations of
both client & server.

The fact is that when one introduces a new media type, it works with
any of the Internet's Network-based APIs (i.e. transfer protocols).  
Examples include HTTP, NNTP, FTP, SMTP, IMAP, etc.    

The key isn't necessarily using "MIME" specifically, it's the
importance of separating operation semantics from media type semantics.


I'd also note that HTTP is increasingly taking over from SMTP,POP3, and
IMAP as a way of sending & receiving email.   There are many emergent
factors to why this is happening, but the properties required to enable
such emergence are arguably due to the architecture of HTTP.

Approaches to Internet-scale computing with only custom operations &
protocols have mostly been successful if a single entity controls both
the implementation of both clients & servers (e.g. Napster, Blizzard's
World of Warcraft), unless the protocol has relatively simple semantics
to reverse engineer & cross-test (e.g. multi-network instant messaging
clients like Trillian).

> >  Really, I am totally not getting your point. Can you explain?
> 
> You talk of standardized MIME types, a theoretical thing, being the

Seems to be working quite concretely for Newsgroups, Email, and Web
pages today.

> "solution", I doubt that we will get to a complete ontology of MIME
> types that is standardised across the globe and therefore REST

MIME types are just identifiers, not ontologies.  To contrast, I don't
think anyone's aiming to create a global ontology out of XML
namespaces.  They're both useful tools for disambiguation via
registries.

> WS-* is further along in the standardisation of industry verticals
> and
> this is liable to accelerate in the next 12 months.  I'm just not
> seeing the business case for them re-doing the effort for REST.

One arguably can create RESTful stanards on top of WS-*, though it's
unclear whether they'll flourish since they often lack cheap & robust
enabling infrastructure to complement the specifications.    

OAGIS, for example, has done this very thing by separating nouns &
verbs in its specifications.   These specifications are on the right
track to being RESTful, though they lack support of a standard set of
processing infrastructure to provide working, freely available, code on
how OAGIS proxies, gateways, etc. would work.  HTTP & HTML had the
benefit of a freely available Mosaic and Apache in its early days.

I don't think the REST camp wants anyone to redo their industry
vertical specs, as I don't think that's their area of interest.  If
industries are successful with such things, that's great.  But,
frankly, many believe (and I include businesspeople in this category)
that such efforts are a waste of time, since they'll inevitably be
subverted by competition. 

As an aside, if anything is true in this industry, there *always* is a
rebuild & respec of the infrastructure every 10-15 years.  The old
stuff doesn't go away, but open source developers come up with new
ideas, academics create new theories, and vendors keep needing to sell
the next shiny thing.    It will take a monumental change in the long
term investment valuation of software and IT assets for this trend to
change.   One can hope....

Cheers
Stu



 
____________________________________________________________________________________
Any questions? Get answers on any topic at www.Answers.yahoo.com.  Try it now.

Reply via email to