Joel Spolsky has a
rant about what he calls the architect astronaut. Simply put, people latch on to a technology concept while missing what makes people intrested to begin with, the features.
At first glance, It sounds like he is whining about marketing speak in technology announcements. As if he didn't figure out that every technology "is going to revolutionize your daily life". He does though name names, and there does appear to be people who try to sell something not because it does anything more useful, but because it is built on some new technology. It's become known as the Killer App dilemna.
The killer App dilema is when a new technology is developed to solve a problem in a technology but has associated costs that are too high (learning, new tools, etc) in relation to the problem it fixes. This problem is made worse by the love of geeks for new technologies. Take the technology I worked on for the last year and half, "QoS" (Quality of Service). The QoS group at microsoft started in the windows media team to solve a need. I assume that need was "Quality Video Streaming". At some point the group moved to windows to develop a soln for the windows platform. A combination of RSVP and marketing was adopted. Its a good soln. Except that it had a deployment problem. Custimers weren't feeling the need for QoS enough to justify the infrastructure costs of adopting it. They choose two things. One, PC's didn't produce any network traffic that needed QoS and should get QoS. Two, known techniques such as increasing bandwidth were easier. The killer app for QoS should have been Real Time Communications (RTC) but the timing was off. Even on a quiet network rtc still had too much latency and quality issues to make qos useful for them at the point of the win2k launch. QoS could do something in the right infrastructure, but it was largly a check box on product information sheet.
Reading that last paragraph over, I got out a rant I needed to but lost track of my point. Typically each of these technologies Joel refers to are improvements in the pervious thing. Yes xml-rpc/soap etc are just another RPC. However it's a RPC that just might finally scale outside an enterprise. Is this detail important? It sounds like it. Does it mean anything to the end consumer? Not directly, but potentially it lets consumers get walled garden benifits outside the walls if the companies and consumer let it.
As for the architects themselves, I haven't met the astraunaut variety at ms quiet yet. Though it reminds me of concept I latched on to when I was in high school and playing with corba. I wanted to create an OS with corba in it's kernel, with no other reason then... it sounded neat. I think today more then ever I understand why that was a really silly idea, not because of itself, but because of the need for motivation, and understanding the implications. Amazingly enough, sun did that with java a year of two later. I'm still unclear on what let them make the buissness descision to work on that, but I've felt the geek pull myself.