RE: Web services are widely adopted ???

Edwin,

Widely adopted is of course a relative term.  Not as widely as XML or HTTP,but more widely than DCE, CORBA, or DCOM I'd suggest.  Meaning available in commercial product and accepted by customers.  It's too early in the cycle to measure adoption wihtin enterprise application deployments - certainly there CORBA is more widely adopted for example.

I agree Web services and REST are very close already, which is one of the reasons I dislike the purity arguments.  We need to be practical, and understand the benefits to business of techniques such as tunneling, and not reject them simply because they don't fit a theoretical idea of how the Web should be used.

I am also not opposed to compromise, but compromise requires each side to accept the validity of the other's view, and I had not been seeing that acceptance in many of the emails going back and forth, basically making the same arguments as have been made for three years, and then stating that the other side doesn't accept them because of a lack of understanding.  That is just wrong and completely unhelpful to the WG's goals.

The other thing we get a lot of on the list is citing chapter and verse, and referring to a list of people who support the same view, apparently for the purpose of helping those who don't understand to recondider their understanding.  Again, this is not helpful to the goal.

Yes, absolutely let's work on compromise, but if what we keep hearing is an unwillingness to consider opposite views on Web services definition and architecture, it will not be possible to compromise and we will not progress.

Thanks,

Eric



-----Original Message-----
From: Edwin Khodabakchian [mailto:edwink@collaxa.com]
Sent: Sunday, May 18, 2003 1:53 PM
To: Newcomer, Eric; 'Walden Mathews'; Baker, Mark; 'Cutler, Roger
(RogerCutler)'
Cc: www-ws-arch@w3.org
Subject: Web services are widely adopted ???


Eric,

I have collected a few snippets from your previous posts.

<eric>
Criticisms founded on purely technical grounds or on the subject of
"architectural purity" completely miss the point of what we need to do.  At
the end of the day, marketplace acceptance is the only measure that matters
for a standard, and the current Web services have been widely adopted.
...
The purpose of my email was to highlight the significance of factors outside
of technical and architectural purity.  Part of the argument I often hear
about REST is that it has succeeded, therefore it's good.  In the case of
SOAP, WSDL, and UDDI the same is true, and that viewpoint needs to be
acknowledged.  Web services products do not implement REST, they implement
SOAP, WSDL, UDDI, and a number of other specifications that are starting to
emerge.
...
Over time, the industry may move more toward the pure document exchange
model.  And SOAP does include a document oriented interaction type that
provides some foundation for this.  But let's defer this step till after
we've worked out the architecture for the current world. Yes.
</eric>

Although I agree with you that adoption is key, I would strongly disagree
that web services have been widely adopted. 

Most of the current use of Web services within the enterprise are RPC
experiments. 

I personally think that RPC will quickly fade away as an aberration of the
web services (the same way, Java applet were cool but faded away) and the
enterprise will instead adopt document and event driven interactions because
they are a far superior solution to the integration pains that the
enterprise are experimenting.

Re: REST versus SOA, it seems to me that the 2 camps are much closer than it
might appear. REST is missing some of the key pieces addressed by XML
Schema/WSDL, WS-Addressing, WS-Security, etc... 

The only place where there seems to be a strong disagreement if the unified
interface. Unfortunately, there aren't enough mature real world deployment
to be able to judge the pros and cons of either approach went it comes to
adaptability. As Ann points out in here email, SOA has some benefits when it
comes to design: I can define one order management portType and group 20
methods. On the other hand, if you look at all existing solutions used for
integration (EAI and B2B), they are all bases on a message bus which exposes
a very simple interface (susbscribe, publish).

I do not think that we should let that difference fracture REST and SOA.

Edwin

Received on Monday, 19 May 2003 06:43:50 UTC