In Modern Operating Systems, Andy Tanenbaum put it best. I can't remember the exact quote, but it was something like:

Running anti-virus software is akin to locking the barn door after the prize horse has escaped.

SIP is a fun feature, but I think it has the potential to be more. I agree, it's "nothing more" than simple collocations, but I'm not sure how that's a critique. In the end, all software boils down to simple assembly instructions, nonetheless the magic emerges.

For further commentary, see this blog post. SIP is just simple collocations, but that's exactly what makes it so appealing. It's moving indexing and search in a more semantic direction, but in a way that's reasonable to design and implement.

This is probably premature, but I've even heard some pundits compare SIP to PageRank.

Amazon has done some interesting research and development lately. In particular, look at Amazon's EC2 cloud computing platform, as well as Amazon's statistically improbable phrases (SIP) algorithm. I have a fetish for natural language parsing, so SIP is particularly interesting to me. These are innovations.

The ISPs will continue down this path until it is no longer economically feasible to do so. And that day *is* coming. One day, it'll be more expensive to play these cat-and-mouse games than to just give away cheap bandwidth, disk space, etc.