Via Stefan, some commentary from the prolific Elliotte Rusty Harold about the WS-Addressing core and SOAP specs recently reaching W3C Candidate Recommendation status. My favorite line is this one:
Besides nobody's actually doing web services over anything except HTTP anyway.
Mr. Harold appears out of touch with what's going on in the enterprise integration space, where folks are doing web services over MQ, JMS, IIOP, and a variety of other protocols, not only because these are the protocols that are actually used in the real-world enterprise, but also because HTTP has issues for certain approaches to enterprise integration. Now, the obvious counterargument is, "Well, those aren't real web services." Even though that's a rather pointless purist argument, I've nevertheless already explained in detail why I beg to differ.
I could write more about Mr. Harold's apparent misconceptions in this space, but Ted Neward has already taken care of it, so I'll just point people (hopefully including Mr. Harold) to my "Toward Integration" columns since quite of few of them cover web services and have already covered this topic in great detail.
Comments (4)
What do you think about specs such as WS-ReliableMessaging in this context? Are they not a first step towards making non-functional features such as reliability independent of the transport, removing one of the reaons to prefer e.g. JMS over HTTP?
Posted by Stefan Tilkov | August 30, 2005 4:17 PM
Posted on August 30, 2005 16:17
Speaking generally and not about any specific protocols or transports, while you can sometimes build reliable messaging over an unreliable foundation, you can't really build true asynchrony over a synchronous subsystem, nor can you build a fast transport over a slow one.
On a more specific level, enterprises can't afford to "rip and replace" their existing underlying technologies every time a new WS-* spec comes along. WS-ReliableMessaging will no doubt eventually help tackle certain enterprise IT problems, but the reliability it offers certainly isn't free, and there's just no way enterprises will use it to replace all the protocols and transports that already successfully drive their IT systems, unless and until they're sure that it makes sense cost-wise and performance-wise.
Posted by Steve Vinoski | August 30, 2005 11:46 PM
Posted on August 30, 2005 23:46
I'll second that - and you didn't have to use a NASCAR analogy ;-)
Posted by Kevin Love | August 31, 2005 4:50 PM
Posted on August 31, 2005 16:50
Too true. Having sat on both sides of the vendor/consumer divide I cannot agree more. I joined an organization whose first foray into WS adopted most of the bad practices. It took almost a year to ammortize some of the worst side-effects. Rip is risk and replace has questionable rewards if it doesn't deliver clear business value. It'll take another while before we can finally go WSDL first. Technically we could bite that bullet tomorrow... but its always prudent to wait until you're 100% sure that the risks have been minimized/eradicated.
When you have web services responsible for maintaing $50m+ or any significant weekly turnover 24x7 you simply can't rip and replace. You improvise, adapt and overcome.
Products like Artix and standards like JBI are a big help for those yet to adopt the angle brackets. In many ways, CORBA is still a better option if you can avoid transacting business over the bare bones internet. It isn't perfect but for the most part it is interoperable. It is well documented, well supported and well understood. It isn't sexy... but it just works. WS-* still has a long way to go before you get the same quality of service out of the box.
Darach.
PS: Steve, don't forget to ping when you 'fix HTTP' :)
Posted by Darach | August 31, 2005 4:53 PM
Posted on August 31, 2005 16:53