This raises an interesting issue. Since this appears a criminal investigation, related parties are unlikely to easily gain access to the materials. For example, in a drug seizure context, the government basically assumes 'too bad' or complicity for related parties/owners (think seizure of an auto or house allegedly implicated in drug crimes). Understandably, technologists see a significant distinction. Unfortunately, the law lags 5-10 years behind reality so while these issues are obvious to most technologists, these issues might not even be comprehensible to some in the legal community. That said, when you are dealing with 1) data servers (often remotely hosted), 2) massive amounts of data potentially unrelated to the criminal investigation, 3) potentially easily segregated, electronic, data silos (in other words, each user has its own dedicated, protected area), and 4) a collective environment, the analogy to auto seizures seems to obviously break-down (more like the seizure of an entire 500 unit apartment building because someone sold crack in #203). Thus, one would presume the affected parties (assuming no complicity--which the government probably will argue is uncertain) would need to challenge the seizures in a court with jurisdiction--which raises its own complexities and costs (plus predicate issues of standing to sue). But, note, there is perhaps another serious issue here related to additional liability. When the servers are seized, the current astounding breadth of government review of the materials seized might implicate others in crimes. The situations is a not-unexpected conundrum (for some of us), and one that you probably won't find featured in your shiny, cloud-computing-will-save-the-world marketing brochures.

Yes. I use both S/MIME and GNUPG email encryption--with GNUPG preferred--to the extent practicable. However, the major problem with encrypted email is the "other half" of the communication. I invariably find that people do not reciprocate with encryption or do not use encrypted email at all. (In fact, in the past year, I can count on one hand the number of encrypted emails that I received from outside.) Despite point-and-click encryption options, despite automated encryption (e.g., FireFox can automatically encrypt), despite freely accessible software, the reality is that most people do not use encrypted email. Unfortunately, it will probably take a major event to help people recognize the volatility and accessibility of email. (And in the US, the so-called Stored Communications Act or state wiretap acts provide little, if any, protection.)

I adamantly disagree. I have completed the Introduction to Machine Learning lectures and course homework assignments. The class, as an INTRODUCTORY course, does an exceptional job of integrating theory and practice. The learning-by-doing methodology helps the student to apply the lecture materials. Yes, perhaps some of the INITIAL assignments appear to result in "minimal code." However, as in excellent writing, paring the assignments to four lines (or an efficient code set) requires significant editing, understanding of the concepts, and familiarity with the prototyping tool (GNU Octave in this case). Thus, the student, initially, applies not only machine learning principles but is also gaining familiarity with an initially arcane (albeit excellent) coding tool. The later assignments do offer more complex coding opportunities and do offer charting, graphing, and analysis. These courses, as is my understanding and I think made clear in the course offerings, are intended to make available exceptional instructors to a wide variety of learners--from persons with no computer science (or even high-level math) backgrounds to experienced learners with an interest in the topic. The courses act as a catalyst for further self-study (for those of us who like an introduction so we can pursue further in-depth study). Furthermore, I think it is unfair to Stanford for others to classify these classes as "Stanford courses." As the materials make clear, these are not "Stanford courses" but __parallel__ Stanford courses. No credit is offered or suggested by Stanford. Apparently, tens of thousands have signed up for these courses. I am ecstatic about future offerings as are many.

University of Wisconsin-Madison A professor mentioned something new called "hyperlinked" documents. An over-the-shoulder demonstration in the computer lab (I seem to recall on a Mac but not certain) was intriguing. The thought at the time, if I recall, was for powerful bibliographic and definitional documents for research purposes. Remember, this was only a simple text document (black and white display) with underlined text that could be "clicked on" to open a new, related document associated with the underlined term. That is, the example I saw would have been a sample article with, perhaps, a word or term underlined/linked. About three years later (c.1996), I ordered my first book from a strange, online bookstore named Amazon .... (Then again, in 1993, that was in the age of using KERMIT to download email list-servs to read at home. Another time ....)

I commend Canonical the community for releasing a Linux distro, but criticism is also helpful (and entirely warranted in my opinion). I started using Linux circa 1997. There have been glitches in Linux distros over the years. However, it seems this version was rushed (these defects are blatantly obvious--e.g., the launcher bar does not work????) and certainly places this distro release at or near the top of the worst-of list. After installing Ubuntu 11.04 (twice) and both x32 and x64 versions, the graphical interface simply fails. First, even with otherwise decent equipment, the so-called Unity interface will not run at all. So... Ubuntu defaults to the legacy, classic GNOME display. Summary in one word: unusable. (OK, horrible or unacceptable would also suffice.) The "legacy" interface has numerous defects making it practically unusable (e.g., the launcher display does not work, windows randomly display the title bar in random inversion on top, right, left, etc. of the window, text randomly goes into "CGA" mode (big, chunky, letters), disappearing windows, disappearing window gadgets. I could go on. The point simply is: Linux is (otherwise) a mature, robust system. Why defeat that legacy with so-called "cutting edge" technologies with little, or no, compelling benefit? Yes, Linux users vary widely, but I believe many use Linux on older systems or simply for a more stable and streamlined system. Yes, some Linux users enjoy experimenting; others want a stable, working basic install and basic usable system. Frankly, this release sets-back Linux five years or more in my opinion. Rather than being able to convince potential users, we are back to the hope-we-can-clap-it-together-just-to-get-the-UI-working mode of yesteryear. Yes, it is fun to customize Linux personally; BUT if you are trying to introduce or use Linux in a more restricted environment, the system simply cannot be as it now is--bifurcated UI and (in my experience) both unusable? (Sorry to the team. No offense meant. Healthy dialog intended.)