It's definitely possible to use php only for full text search.
There is a PHP only implementation of Lucene that let's you update it's index in php and that you can call inline or periodically via cron for example.
Furthermore, you don't need a custom build of mysql for sphinx to work, sphinx runs as a deamon and works with a standard version of mysql (at least standard ubuntu packages of mysql and sphinx). Sphinx works really well for the cases I've worked on.
@Tim: If you are talking about Zend_Search_Lucene then its indeed an option, but not a great one. We have used it for a project at Liip and it did the job, replacing a Google Search Appliance no less. But its not really a great alternative. That being said its certainly something to look into supporting as well. It really is just a matter of people stepping up to work on this. This is really the great thing about PHPCR, the API allows total freedom in terms of libraries used to implement it.
As for Sphinx, this is great news. I wasn't aware that it can also hook into MySQL via the Daemon, but I guess this is now possible due to the plugin API in recent MySQL versions? Again it would be great if someone contributed an implementation for this.
Keep up the good work! You're doing great. I know I haven't been able to get my team on board to help develop the CMF further, but I'm still trying. Articles like this are very helpful to me to show my team where you're headed and where you've come from so they know you're not trying to box them in, but give them tools to do their job more efficiently and more standardized.
I think the PHPCR is awesome and should be used on any project that has content, especially if this content might be used in multiple "applications". I'm really looking forward for some time to build some COPE Bundles on top of the PHPCR to pull out syntax and ensure a separation of content and markup.
Again, keep it up. You're doing a great job!
Jarvis (aka BallisticPain)