In August I posted an issue on the Zend Tracker in regards to a noticeable bottleneck with serialization. ZendAMF performs at least 5x slower than AMFPHP, which is jarring when retrieving a high volume of rows. Many just shrugged it off and stayed with AMFPHP on more demanding data-retrieval projects. Today Mark Reidenbach contributed a patch; diff files. ZendAMF wasn’t using referenced variables, and was also sacrificing speed for size by checking for strings using array_search. The size difference resulting from the changes on his project are almost double, but still only kb: 179kb to 287kb. So pick your demon. Mark reports that these two changes have brought serialization time back to roughly what it was with AMFPHP. Big ups Mark, I salute you and will buy you a bucket of beer at the next RIAdventure !!
You can grab the .diff files here:
http://framework.zend.com/issues/browse/ZF-7493
Yes, I also go by Prabhu Singh, a name given by Yogi Bhajan
