On Fri, October 27, 2006 6:25 pm, Niklas Edmundsson wrote: > I would have been most happy if this had been fixed ages ago so I > hadn't been forced to spend lots and lots of hours kludging stuff > togehter. At least, my kludges seem to have sparked some development > in this area, so they have served some purpose other than enabling a > non-profit computer club building a FTP/HTTP server that actually > works.
When Brian Atkins stood up at Apachecon and presented a talk pretty much summarised as "we don't use Apache's cache because it sucked, and here's why", that was a wake up call to realise that the disk cache in its current form sucks. We have significant contributions from two people - Davi Arnaut and Niklas Edmundsson, and I've been integrating the issues fixed by both these contributions into a coherent workable whole, so that the effort spent by these people isn't wasted. Both of their efforts have focused on different aspects of the cache, making this workable. Some parts are not RFC compliant, other parts are not implemented elegantly enough, but these are details that need to be raised, addressed and fixed, not used as a feeble excuse to abandon the effort and return to some cache code that nobody wants to use. I see lots of comments on the code, but the comments are summarised as "the cache is fine as it is". It isn't. If it was fine, key users of network caching wouldn't be standing up saying they're using something else. Regards, Graham --
