W3C home > Mailing lists > Public > www-ws-arch@w3.org > January 2003

RE: REST; good for humans and machines

From: Newcomer, Eric <Eric.Newcomer@iona.com>
Date: Sun, 5 Jan 2003 14:45:34 -0500
Message-ID: <DCF6EF589A22A14F93DFB949FD8C4AB2916F6A@amereast-ems1.IONAGLOBAL.COM>
To: "Baker, Mark" <distobj@acm.org>
Cc: <www-ws-arch@w3.org>

Mark,

If I understand your reply correctly, you are disagreeing with a point I was trying to make because of a technicality of the language I used, which is actually the point I'm trying to make.  English is not a compliable language, and what I meant by what I said is different than how you replied to it.  

>Again, that's a common mistake of CORBA folk, and Web services folk; you
>can't compare IIOP and HTTP.  IIOP is layer 6, HTTP is layer 7.  They
>are entirely different beasts.  It's like saying "Ethernet is better
>than IP", or "IP is better than TCP"; it doesn't grok.

It doesn't matter if IIOP is layer 6 or HTTP is layer 7.  I know it matters to you because that's how you think about them.  But what really matters is what you are trying to accomplish, not what layer in the stack some spec writer put things (and please don't tell me that whoever decided on the original OSI seven layer stack got everything exactly correct, or and that there were not tradeoffs or ambiguities in that spec, or in HTTP or any other spec for that matter -- again, English not being compilable into binary meaning it's impossible to have an objective test for any individual human's interpretation of a spec).  

This discussion reminds me a bit of an argument I used to often have with a former colleague while working on a spec draft.  He would write text that no one else could understand, and when I pointed that out to him, he'd say "it's better to be precise than understandable."

A debate like that is very hard to settle ;-).

So I'll just try once more to rephrase in the hopes of being more understandable than precise: people do compare HTTP and IIOP despite the fact that they are, technically speaking, defined at different levels of the stack.  And they do so because what we are trying to accomplish is the same, or similar enough, to warrant comparison.  

Actually, by your logic, the comparsion of REST to Web services doesn't make sense, either, as they also are defined at different levels of the stack.

Eric

-----Original Message-----
From: Baker, Mark 
Sent: Sunday, January 05, 2003 2:32 PM
To: Newcomer, Eric
Cc: www-ws-arch@w3.org
Subject: Re: REST; good for humans and machines


On Sun, Jan 05, 2003 at 10:02:20AM -0500, Newcomer, Eric wrote:
> I think that if the REST folks had designed Web services, they would be different, but then again, HTTP-NG was different, and so was HTTP-EF, and neither gained widespread acceptance.  Nor did ICE, XML-RPC, and about a dozen other XML protocol proposals.
> 
> So yes, certainly I think everyone can agree that theoretically you could design and implement pretty much equivalent functionality using HTTP without SOAP and WSDL, but the larger, and very much more difficult problem is how do you ensure market acceptance?

I thought I had been clear on my position here.  "REST folks" would not
have designed Web services, because we don't need them.  We can use REST
to solve the same problems that people are solving with Web services.

> I know that many of you remember very well about three years ago when the adoption of SOAP was very much up in the air.  And many of you no doubt remember the COM vs CORBA wars of the mid-90s.  If the industry had been able to agree at that time upon a single RPC standard, CORBA would have been much more widely adopted and therefore more widely successful, perhaps even becoming the standard for "Web services" as many felt it should.   (And by the way, CORBA folks think anyone who'd use HTTP for any kind of distributed computing is nuts,

I used to be a "CORBA folk".  I did *heavy* CORBA work from 94 to 98.
I later learned that HTTP was a better way to build those systems
(large scale telecom network services management + integrated
billing/workforce/sales/network).

> and I'm sure we could drum up an equally endless religous argument about why IIOP is superior to HTTP and should have been adopted, as was once proposed, as an Internet standard. Then we wouldn't have all these issues with HTTP...) 

Again, that's a common mistake of CORBA folk, and Web services folk; you
can't compare IIOP and HTTP.  IIOP is layer 6, HTTP is layer 7.  They
are entirely different beasts.  It's like saying "Ethernet is better
than IP", or "IP is better than TCP"; it doesn't grok.

> One other thing, although I know this is probably asking for too much -- it would be nice if we could also acknowledge that there are no "objective facts" in this discussion since by definition we cannot understand what we don't understand, and every human brain, as every human, is imperfect. There is no absolute, objective truth since to know it we'd have to be non-human.

For the most part, except for that misconception mentioned above.  8-)

MB
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca
Web architecture consulting, technical reports, evaluation & analysis
Received on Sunday, 5 January 2003 14:46:06 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 July 2007 12:25:12 GMT