Jakob Voss wrote:
Maybe diffs are not the most important performance issue but surely they can be speed up.
Yes, it would certainly a useful project if anyone wants to work on it. The gains would be comparable to any other performance project on this scale. The C++ module could do with some optimisation, and the PHP equivalent will continue to be used by external users if not by Wikimedia. And they would be useful projects for standardisation and submission to PECL and PEAR, IMHO.
-- Tim Starling