OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help



   RE: [xml-dev] REST as RPC done right

[ Lists Home | Date Index | Thread Index ]

From: Paul Prescod [mailto:paul@prescod.net]

>You said the other day that the essence of hypermedia is nodes and
>links. If that's so, then what problem domains do not naturally lend
>themselves to hypermedia?

No I didn't.  I gave it as a quote for a definition.  What I said 
was that without a definition for distibuted hypermedia, Fielding's 
paper has a hole in it.

>> Of course you don't.  And we haven't.  You are
>> winning a lot of converts out here, but the code
>> is on the loading docks and it is not as Leigh's
>> article suggests, that we are content to let it
>> fall apart later, but that just as we in the SGML
>> industry had to face up to the reality of HTML
>> ubiquity, we may not have a choice. 

>HTML did its job well. 

You said it was the awful part.  But I don't disagree 
with it doing a job well.  When it comes down to creating a vocabulary 
that thrives in some niche, gencoding wins hands down every time.  Ask 
yourself why and the answer is the same reason as the fixed methods of HTTP: 
it's very predictable.

>One of the inventors of SOAP says he doesn't know if RPC will scale up to the size
>of the Internet. 

It may not have to.  Perhaps it only has to work for the keiretsu 
of industries that care to use it.  To me, the thing that would 
be a big improvement would be two way persistent pipes, but given 
the laws that govern large networks, I don't know how reasonable 
that is so I will settle for coordinated networks of typed 
transactions broadcast from central servers and received by 
qualified listeners.  

>Most people on xml-dev seem to think it won't. 

The jury is still out on that.  Again, code is on the loading 
docks, there are success stories for frameworks such as dot.net 
and just as we had to subset SGML to fix the mess that HTML 
left in its wake, someone may have to recombine HTTP-like 
thinking with RPC-like thinking to make .net work.  Perhaps 
only some processes are RPC-like and the rest are just 
good old fashioned HTTP string schlepping.  Hybrids are 
often the more adaptible design in nature.   What will make any 
of these go will be adaptibility to a changing environment. 

>Who is left as a true believer? The RPC bit of SOAP won't work right and what's
>left when you take that away is too generic to be interoperable.

Ummm... you may be right but I don't think so.  I think SOAP will work 
for those that buy frameworks and make it work.  I think we will 
have some trouble calling that "the web" although they will be 
Internet business applications but that is because the defintion 
of "the web" is quite conservative.

>It boils down to this: Tim Berners-Lee built the world's most successful
>and widely used information system. 

Ummm.... Tim put a hypertext on top of 30 years of work 
by DARPA et al.  Then a lot of people have had to work like heck 
to improve it so it was more useful than emailing a .txt file 
to one's colleagues. Gavin is right about the revisionism. 

Yes, a quick and dirty HTML was better than a perfect 28001 when it 
was time to give it to mom and pop to use and that is why it succeeds.
Gencoding always works until one has to adapt it, then the stresses 
begin to out.  Just as true in 1990 as it was in 1970.  
But to make it be anything more than click and get, a lot of other 
people had to add their own bits to that from their own life's work.

>Everything I have ever read by him says that he believes that the secret 
>is the universal namespace. 

He does.  A uniform namespace is what makes ANY information *system* work 
uniformly from a C function call to UPS.  Duh.  Otherwise, we get  
yeOldeBritish mail system; lot's of local address types that must be  
interpreted locally.  Inefficient, but they still deliver the mail to the 
Queen fairly quickly, and it ramps down from there to Welling.

Why can't a standard computer and standard Internet succeed as the 
TV and automobile did?  Because no one uses a TV to design a TV 
or an automobile to design an automobile:  feedback, relentless feedback. 
The only question is, are the systems evolving or merely orbiting 
a small predictable set of attractors at this point in their history?

Hubris is believing it must be one system to succeed.  Given the rampant 
spam and the irreversible loss of privacy, perhaps a bit of non-uniformity is 
to be preferred.  Sorry, Paul, I just don't buy into that universal vision.  
It reeks of the pit.  Put down Turing and read Milton.

But uniformity is always useful for a period.



News | XML in Industry | Calendar | XML Registry
Marketplace | Resources | MyXML.org | Sponsors | Privacy Statement

Copyright 2001 XML.org. This site is hosted by OASIS