There are also other ways that the product screws up user interaction with Web applications.
On another note, I’ve read a number of somewhat paranoid theories about why Google has created this product, but I haven’t read anything that really makes sense to me. The idea of Google keeping a log of every single Web site I visit is a bit unpleasant, though, especially when it’s theoretically possible for them to tie that not only with my Google searches but also my email inbox. I’m not anti-Google by any stretch of the imagination, but I think it pays to be mindful of how much information you give to any company.
Update: James Governor has an interesting post on viewing Google as a measurement company, which certainly dovetails nicely with the side effects of providing proxy servers to the world. I had forgotten that Google purchased Urchin, a Web analytics firm, recently.
Another update: Tristan Louis points out that Google’s Web Accelerator may be built to help Google grow its index. Very insightful.
Another update: Getting back to the issue of the prefetch features following links that alter or delete data, the real problem is misuse of the HTTP GET method. This is a problem Sam Ruby warned people about some time ago.