Surprisingly, out of all the comments, this one caught my eye. I've been hiring programmers since 2005 that have to maintain or extend PHP systems. In 2005, it was difficult to find programmers that knew only PHP (or Perl, which was still used) who had any familiarity with object-oriented programming. Over time, PHP has bolted on features, and at each point, the vast numbers of available PHP programmers are (understandably) not familiar with those features. And obviously the systems that have to be maintained don't have these features. So the armies of PHP programmers are well matched to the old systems. This eliminates a huge risk to the business -- finding relatively inexpensive developers. This is an often cited advantage of PHP. But only the PHP of 2007 (or so) is relevant.

The PHP of 2009 might be worthwhile for the creation of business value (and possibly the reduction of maintenance-related risks), but as I said, the easily available programmers have no experience in those features or concepts. The more common scenario is that a particular programmer has experience in some other language that already included those features and doesn't mind working on old PHP systems. Believe me, there are not legions of these people.

And regarding the 'PHP of 2012', the major problem isn't the availability of programmers who have experience in the new features (they almost don't exist). It is the availability on distros stable enough to support an SLA or provide a modicum of security assurances. This problem applies to all languages though. I'm just pointing out that the 'PHP of 2012' is not as much of an option as it may appear.