I was talking to a very large financial-sector company in the middle of America in the middle of last week, and the cognitive dissonance around this whole WS-thing is palpable.
These particular guys, they’re a classic hybrid shop with a Java ranch and a .NET plantation and everything needs to talk to everything, and now they want “Web 2.0”, by which they mean interesting community apps built fast. They’d just had their big Microsoft corporate visit, and the word, from Ray Ozzie on down, is that it’s all WS-* all the time.
But they’re cautiously interested in modern Web tech like LAMP & Rails, where the support for that kind of stuff is mixed at best.
Then I come in and give them the REST message and take a few whacks at particularly soft WS-targets (no shortage of those) and talk a little APP, and everyone gets real uncomfortable. I can tell you what we need; some packaging and labeling. Something like the Atom Publishing Protocol, it’s so easy to understand what it does and why you might want to use it, there’s lots of comforting implementations, and when you say “That’s REST at work”, well, heads nod. REST needs a few APP analogues in other sectors, just to help make this stuff easier to understand.
Damn, it’s been a long strange trip. You hear sad stories like Mark Baker’s, and it makes me feel guilty; I’ve been calling the WS-emperor naked almost as long as Mark, but from the comfort of a safe corporate designated-radical job, not really running any risks. We all owe Mark a vote of thanks and, hey: he told you so. And after all that, it’s perfectly OK for there to be anger: “it’s hard to understand how the same people who have screwed up so much of the past near-decade should be given the space in reputable publications to make these types of statements without them being attached to a letter of resignation.”
And yeah, the guys in Redmond can chant their WS-mantras and we’ll even build things like WSIT so Javaland can talk to them. but, you know, the LAMP guys mostly won’t. When the CTO of HP’s Business Technology Optimization group asks, in Is REST eclipsing SOAP?, “Are standards like ATOM or APP eclipsing SOAP? Yes, definitely. For very simple reason—their adoption is way broader and tools support way better”, it’s starting to look like Redmond’s located on a WS-desert-island.
Contributions
Comment feed for ongoing:
From: Karl Waclawek (Jan 29 2008, at 10:56)
Why is it always WS* vs REST?
If we are talking distributed apps (and not just web apps) then there are other choices. I have been using ICE (http://zeroc.com) quite successfully, in one instance replacing a SOAP web service where Java clients just could not interoperate with it, as they required adherence to WS-I.
Karl
[link]
From: Lance Lavandowska (Jan 29 2008, at 12:58)
Yo Tim, if "middle America" is code for Minnesota, let me know next time you're going to be in town; let's hook up for some brewskies.
[link]
From: Robert Hook (Jan 29 2008, at 14:37)
"...they're a classic hybrid shop with..."
That rang bells for me, and I would like to put forward another proposition as to why REST/APP (or insert any other pair of acronyms you like) are not widely adopted in the Real World (tm).
My experience (as a developer) over the last two-and-a-bit decades is that higher levels of the organisation read a magazine / web page / powerpoint presentation, are converted to a new religion, and sweep down from the mountain carrying stone tablets. "Thus it will be for evermore." they say, and the developers in the trenches shrug, start delving (for months) through the (scant) documentation, begin the process of getting a product out. Months later, the managers go to another conference, and the whole thing starts again.
I'm in exactly that boat. My company decided, some four or five years ago, to move to an architecture that roughly resembles what later became called "SOA" (remember SOA? It was the Great New Thing To End All Things about three years ago). And we've been building it. .Net. SOAP. RPC. J2EE 1.4 style web services. And there has been about four or five years work in that, and several more to go. Now, of course, our managers want Web 2.0 and Flash and a plastic rocket and a pony.
The trouble is, we have to finish what we have started, and we can't just dump everything and start from scratch.
Thus, I posit a problem from the Real World (tm): the project timelines for non-trivial environments are longer than the period of introduction of The Next Great Thing To End All Things. There are lots of people out there building systems nodding and saying "Yep. REST is better than SOAP. Might look into that in a few years when we get this one finished".
[link]
From: Tony Fisk (Jan 29 2008, at 15:57)
A lot of this resistance comes from people not being prepared to admit that what they're doing isn't the best solution (after all, *they* chose it, or believe they did!).
<i>If we are talking distributed apps (and not just web apps) then there are other choices.</i>
I think 'choice' is the underlying issue here. If you like ICE, great. I'm happy with the LAMP and REST approach. I suspect that the WS crowd do not like either of our attitudes.
[link]
From: Joe Cheng [MSFT] (Jan 30 2008, at 11:34)
While the sales pitch of a MS sales team sets you up a very nice strawman, it's just not true that "the word, from Ray Ozzie on down, is that it's all WS-* all the time".
The current version of Windows Communication Foundation was explicitly tailored to support REST. ADO.NET Data Services (f.k.a. Astoria) is based primarily on AtomPub. The multi-master sync technology that Ray Ozzie personally worked on in 2005 is based on Atom. And thought leaders in the company like Don Box, Pablo Castro, Yaron Goland, and Dare Obasanjo have long since given REST its due credit. Those are just the projects and people I can think of off the top of my head. (BTW, I hate myself for using the phrase "thought leaders" without irony.)
Knowledge of the virtues of REST may not have filtered down to every corner of the company, but it's not because we're all still drinking the WS-* Kool-aid.
[link]
From: Mike Champion (Jan 30 2008, at 13:39)
Nobody, and certainly not the numerous web-facing folks at Microsoft, would dispute that there are lots of soft targets for REST out there, and WS-* is massive overkill for them. Mark Baker deserves immense credit for pushing back on the WS-* herd mentality in 2002 or so. But the battle against SOAP on the Web (for ordinary public services) was won long ago, and RESTian ideas have taken hold and proven themselve in the WS-* world as well as the Web. Most folks in the web services world have taken a "best tool for the job at hand" stand, but folks such as you and Mark still portray this as a Clash of Civilizations between REST good and WS-Evil.
I'd like to hear about real success stories for "pure" REST-over-HTTP when the targets are hardened -- where protocols other than HTTP are part of the mix, where information is valuable enough to protect against sophisticated intrusion attacks, where end-to-end message reliability is critical,etc. It is quite true that WS-* is not necessary or sufficient in these scenarios, but what about the alternatives? I suspect it will be a bigger challenge to build efficient, scalable, secure, etc. apps without the composable services approach to combining these cross-cutting aspects, but I have no data to compare the well-known WS-* challenges against.
I do believe that most of the real problems getting WS-* systems to interoperate do not go away by taking the SOAP stack out of the picture: you still have the intrinsic issues of requirements imprecision, schema evolution, semantic mismatches, and business competition that make application interoperability fundamentally hard even when there is agreement on architectural principles. One can fault the web services advocates for raising expectations that these would somehow magically go away, but it would be equally dangerous for REST advocates to ignore these matters.
So, how about a pledge to highlight an enterprise-scale REST success story against a "hard target" for every kick at a WS-DeadHorse?
[link]