Some of you may be over looking the filed date of the patent. December 30th, 2004. That's enough time for a lot of things to change.

I do agree that the patent office needs an overhaul though.

than Sun had any documentable plans for such a beast.

U.S. patents go to the first who had the idea, not the first who implemented it.

Um ... Sun did this last year how can google patent it!?

http://www.betanews.com/article/Sun_Unveils_Portable_Datacenter/1161116189

I have designed and run large computer centers, but I don't really understand why this idea is so appealing. the premise is that a site experiences a massive but brief increase in computing or storage demand. how often does that actually happen? it would have to be for significantly less than a year to justify not simply building the appropriate machineroom. a whole c-van full of racked computers does indeed involve less plug-wrangling on delivery, but the same amount of wrangling has to happen _some_ time. currently any significant-sized install will arrive on preconfigured, skid-mounted racks, which involve only about 1/40th the wrangling of bare nodes. a c-van is probably good for about 10 such racks, but wrangling the racks is not any kind of big deal (couple guys for a few days.)

maybe the idea is that you'd take permanent delivery of a containerized cluster, and just build a slab with roof, power and chilled water. I guess that could be done, but it seems quite dubious to me, cost-wise. also, PHB's love to be able to point at and/or tour an impressive machineroom. something that looks like a freightyard doesn't have the same appeal...

It is also a good example of why the patent office needs a major overhaul

It depends on what the specific claims are in the patent. If Google made the mistake of merely patenting a datacentre-in-the-box, that patent could, and will be, readily challenged on the basis of prior art, if nothing else, so why would Google go through the expense, and time, to do so?

Their patent must contain something new and unique to be worth anything.

Unless they believe they can just outspend opponents to get a favorable decision, which is bad judgement on their part.

I've heard subdued voices whispering about IBM support contracts and some hidden location from where a complete mainframe could be heliported to any point in Europe in under six hours - ready to be plugged into the customer's infrastructure and pre-configured with all the customer's parameters.
Isn't that enough for prior art ? If it's true, obviously, and not just some enthusiastic elaboration of an AS400 support contract (less impressive, no need to heliport, but still rather important).