WS-* services and REST services are not competing, they are complementary. Web Style serves another purpose than SOA. Consumer-facing services have other QoS requirements than high-volume, cross-platform A2A transaction services.
That’s really interesting from my POV because it shows that after many years of a debate, that we – the Web folks – still haven’t successfully gotten our message across.
That message? That these services are competitive.
It’s as Sean says (channeling some unnamed prophet);
A complex system that works invariably can be traced back to a simple system that worked
Meaning, in this context, that if you want to build something suitable for “high-volume, cross-platform A2A transactions”, start with the Web and build up. Web services don’t do that; they strayed from that the moment they confused transfer with transport.
Yes, that’s my gmail address. Apparently, a lot of people disagree.
I get a handful of “To initiate the process for resetting the password” emails a week, initiated by M Bakers who’ve forgotten their account names. I also receive a surprising number of emails to other M Bakers. In fact, last week some one of them tried to forward something to his own gmail account and missed! Dude!
And just now, I got a USPS shipping service notification to a Michael Baker from somebody who probably guessed an email address. Sigh.
Two years ago this past weekend, I thought I was lucky to get one of the first non-Google-employee accounts, and to have pretty much free-reign in picking my account name. Alas, ease of rememberance is not without its costs. Oh well, at least I’m learning new stuff… like how to organize a bachelorette (I had no idea!).
Dave Linthicum asks;
With lack of interest in UDDI there seems to be a need for another directory standard to come up and take its place.
As I previously pointed out, directory standards simply aren’t required because we have the Web. But whether or not you buy the “Web for machine-to-machine integration” position, I think it’s pretty clear that because UDDI registries were accessed primarily by humans and not automata, that it was directly competitive with the HTML-based Web.
And that’s exactly what we saw happen; there was the emergence of non-UDDI registries such as XMethods which primarily presented an HTML/HTTP interface, but more interestingly, we saw the UDDI registries themselves being used almost exclusively (from what I heard) through their HTML/HTTP interface.
So what will happen when automata want to start doing actual dynamic discovery and integration, and we need more than plain old HTML? Will they incorporate the lessons from UDDI? Will they try to reinvent it? Or is a centralized registry, or some component thereof, even necessary?
I hear Phil Windley has some ideas in this space. In fact, he and a student of his, Tom Warne, will be presenting this work on my Dev track at WWW2006 in Edinburgh next month. If this interests you, we’d love to see you there.