The web communication that occurred after the Tsunami could make very few assumptions about the reliability of the network – even the assumption that there wouldn’t need to be at some point someone hand-copying a file on a piece of paper and taking it to some other computer because the information involved was critical but the physical infrastructure was no longer there.

RESTful solutions work in situations like that because the computing world is heterogeneous, because often times the most important information being passed is that which says that a child has been found alive, or that survivors should gather at such and such a location, and you are dealing with ancient systems that don’t have the latest in MS-WS* but do have a workable web server stack.

If I have a choice between acting as a node for sending an RSS feed (one that may not even be “safe” XML) knowing that it will likely get there or have the relevant messages get “stuck” in some server because they didn’t install the latest service pack upgrade, then yeah, I’d definitely take that RSS feed.

I think that is a big problem with much of the thinking coming out of Redmond, though I also agree here that the recognition that SOAP is a niche technology (and that all programming is niche programming) is welcome. There tends to be this assumption that complexity adds to reliability, when in fact the more tightly focused a given piece of technology becomes, the less reliable it usually ends up being. This is a lesson as true of programming as it is with the argument of Ferrari’s vs. VW Beetles. A Ferrari is a performance driver’s dream, but it was so optimized for performance that they are in the shop as often as they are on the road. Nobody was likely to be doing high speed racing in a VW Bug, but they were easily the most easily maintainable car on the planet at the time – with this factor, as much as their affordability, being the reason for their general popularity.

Most people don’t NEED the power and performance of SOAP enabled web services and the WS-* stack. More power to Microsoft to recognize that there IS a need there for those that do, and that they are likely to be the customers that Microsoft is most interested in pursuing because they are also the wealthiest, but don’t expect the rest of the world to greedily jump to SOAP because of the expectation that they are more “reliable”.

Fully agree, and sorry that I phrased my piece in such a way as to create any confusion about this. I was trying to make the point that in those situations “where lives are at stake” AND the reliability, transactionality, security, etc. of web services significantly affect the stakes, that might be a better niche for WS-* than for REST. Integrating public safety systems might (I don’t claim to know this domain!) be an example: Would you feel safer knowing that your local 911 system communicates with your local police, ambulance, fire, etc. agencies via SOAP+WS-ReliableMessaging+WS-SEcurity, or via a bunch of POX files that passed across legacy boundaries with GET and POST + a lot of application-level code to route and secure the messages? I’m sure Mark Baker can sketch out a “gorgeous” scenario for how all this MIGHT work RESTfully , but AFAIK the real world developers in that domain are betting on WS-*.