"Redistribution of source code must retain the above copyright notice ..."
Or how about the Apache license:
"You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent, trademark, and attribution notices from the Source form of the Work,"
Maybe the MIT license:
"The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software."
The potentially interesting question is whether the iPad is actually good for non-consumption tasks. All the faculty in my research group purchase Apple products, and most of them are wandering about with iPads. They're all quite happy to use them for reading papers, reading email, etc. However, none of them, despite having been using the devices for some time (close to year for some) even feel comfortable using them for technical email, let alone writing lectures, presentations, or papers.
Again, I don't know whether the software to make those tasks workable _could_ exist; I simply have experiential evidence that it does not yet.
Maybe multitouch will make all kinds of exciting differences. Maybe it won't. As long as you agree that you can't point to any of these exciting results yet, we're on the same page.
Being a practicing mathematician means generating and publishing results. I am interested to see your iPad app for diagonalization results or computability proofs.
As far as I can tell, the threshold for joining sales engineering there is surprisingly low.
BUT... keep in mind, they intentionally don't educate their pre- and post-sales folks about the actual capabilities and flaws in their approach. The training I got was basically useless---a week of having someone read the getting started guide to us.
Apparently, one of the earlier sales engineers had written a "real guide to IDOL" or some such, intended to help new sales engineers actually know what to do. Autonomy's CTO had all copies of it deleted, because it referred to bugs and shortcomings that they'd signed contracts claiming didn't exist.
As corroboration, my first assignment at Autonomy basically amounted to writing (on-site) a component that had been sold without the existence of a single line of code.
I think the obstacles for a startup wouldn't come from Autonomy's core techology---more the amount of stuff they've acquired around the edges. Pieces like KeyView (which is fairly good at reading just about any file format you care to mention) make the whole Autonomy package much more attractive than their search tech would be on its own. Plus, as you said, you have to be willing and able to compete with their sales force, and that seems like a soul-selling endeavor from the beginning.
Autonomy is very good at selling demos for prices at and above what you're listing there.
I spent some time there in consulting, which amounted to attempting to convince customers that the demos they'd bought actually existed. In general, the demos were mixes of Perl and JavaScript that barely held together for the length of the demo, let alone scaled. Autonomy did have some internal large-scale success stories, but they tended to involve heroic (and large uncompensated) work after the sale and customers who were willing to repeatedly threaten to sue until said work was done.
PTX seems quite reasonable to me, but I don't think it's really intended to be a programmer-facing abstraction. NVIDIA wants you to target CUDA, and they'll (try to) ensure that CUDA has sufficient performance advantages to make it worth your while to do so.
Even NVIDIA isn't crazy enough to try to get everyone to use the same ISA again; they're trying to get you to buy into their compiler chain, not into the architecture of a particular generation of GPU.
Barriers are slow, and the more cores are hitting the barrier the slower it will be. The best news on that front was from the new GPU architecture parts, where they admitted that they're still going to expose the relaxed consistency memory model.
IBM's algorithm scales approximately linearly in the number of qubits. So, you'd need more than a few more qubits...