Over at Geomblog, Suresh Venkatasubramanian argues strongly against making post-doc experience a requirement for an academic job in Computer Science:
I think that this is a dangerous trend, for the following reasons:
Quickly, doing a postdoc will become the norm, rather than an option, when looking for academic jobs. I think this is unnecessary from a training perspective for everyone (though it might be appropriate for some).
One of the things that has kept CS viable academically is that people can leave after a Ph.D, go to industry, and still make it back into academia. This no longer seems to be true in places like the natural sciences, with long postdocs. I wouldn't want their career path.
Postdocs are glorified free labor for PIs. Salaries are miniscule, and competition is fierce. And again, it's not entirely clear that fresh Ph.Ds are so incompetent that they need 5 year postdocs to be ready for a faculty job.
Ph.D training suffers, because "you can fix it in the postdoc". I don't think that's healthy either.