Frankly, I can't imagine that even the less prepared script kiddie wouldn't keep all their hacking data inside a TrueCrypt partition allowing him to claim plausible deniability.

That, an open wifi, then claim "it came that way, or I couldn't make my netbook connect, so I had to open it".

Given those basic security measures, what evidence could the police use to incriminate him? Video/screen surveillance? I can't think of any other way.

Aside from what other readers have noted (that getting a 5% return is very optimitics right now), you are not counting long term savings.

That is to say, today we have 1st generation power units which cost 3,500,000$ and save 100,000$ a year. Investing that money in the bank will return you 175,000$, all right, 75,000$ more than buying the machines.

But if you buy the power units, you are investing in power unit technology. So if they get a lot of customers, assume in 5 years time the units will cost 1.000.000$ and save you 300.000$ a year, a 30% ROI, no bank can compete with this. This is what happened with computer hardware, just look 20 years ago!

I guess previous work had the idea right, but actually building a system which can handle millions of links and reply in no time is not a small feature.

This reminds me of the discussion we had previously about the gap from research prototype transistors to having factories actually deliver them.

This is nice in theory, but in practice it has very serious security & management implications. You better don't allow programs to replace its code when called from a normal user, it creates a hell to support.

While I think Microsoft is right with its release cycle, the article is based on the fact the every other browser vendor is releasing snapshots.

For me, the biggest picture is interaction and strategy, not builds. In Webkit, Gecko and Presto, if you are a web developer, you can interact with the engine developer. They have mailing list, good bucktrackers, and a *good attitude* towards fixing bugs.

For Microsoft, if you are using Linux for development (a pretty common case I'd guess) you cannot even try. I doubt Windows users do fare any better. By the way, Windows 7 is not bad, but not usable yet.

We had the same problem, so we went with a different setup. We have two big Debian GNU/Linux servers and a lot of HP thin clients.

Each of these runs at 15Watt, so 10 users consume the same that 3 workstations.

I don't see relational databases going away any time soon.

Most (>70%) of the web is using them, and so far, they've worked very well.

What is missing is good support for them from the programming language point of view.

The nature of relational databases is declarative, as you define mathematically what you want, not how. That's a job for the database, and they've got huge compilers and optimizers for it.

Of course, the SQL language is a leaky abstraction of the pure relational calculus, and you have to know certain rules in order you query can be answered efficiently.

SQL doesn't fit well in imperative languages, where all you can do is write down instructions. Compare that with a language like Prolog, which is OOTB a relational database.