1.  Current popular press sentiment on space shuttle Columbia, from MSNBC.  The most interesting comment is from one of the astronauts, who expressed "disappointment" that (if I apply my own interpretation of his statement ... :) the engineers on the ground fucked up again.

2.  The Millennium Prize problems in mathematics.  Basically, if you can solve one of these uber-math-problems, you win $1 million (and, one might reasonably assume, fame, a tenured professorship somewhere fancy, etc.).  I found this via an article in the August "Communications of the ACM" which argues that one of reasons undergraduate enrollment in Computer Science has dropped off precipitously is that we've done a bad job of making it sound as interesting and important as it really is.  The example here is that one of the Millennium Prize problems, P vs NP, could generally be appreciated by a high school student (it is suggested that the other six problems are not as accessible).  The importance of NP theory is pervasive.  For example, the modern cryptographic techniques used to secure an online purchase were developed following early work in NP (the specific example being the lack of a general polynomial time solution to factoring big numbers).  Further examples of the cross-disclipline impact of CS research include gene sequencing, modern particle physics, and all of the stuff that makes Google possible (or useful).  High school students can certainly grasp the relevance of these advancements, and yet are rarely exposed.

3.  Speaking of Google, the August CACM also pointed to this paper, The Google File System, as a popular download this year, so I decided to check it out.  A couple of cool points:

• The f/s is optimized for storage of large files.  In fact, the chunk size, or smallest storage unit requested and replicated across nodes, is 64 MB!
• Data flow is decoupled from the transmission of metadata.  So much so that the two can travel from the client to primary and secondary storage nodes in different orders, allowing data flow to be optimized for link speed and latency.  Metadata flow is instead optimized for fault tolerance and consistency.