From: Jody G. <jga...@re...> - 2007-06-20 16:22:32
|
Martin Desruisseaux wrote: > In my "history fix" round, I restaured BufferedAuthorityFactory as it was before > ThreadAuthorityFactory and as an independent class (not a subclass of > ThreadedAuthorityFactory). This is because I feel that ThreadAuthorityFactory > could be a subclass of BufferedAuthorityFactory and just override the > 'getBackingStore()' method. We may talk about that later - it doesn't impact > your development in short term on ThreadedAuthorityFactory. > I got no time left for talk martin -- That is why we had the proposal process :-( I will be on IRC if you want to chat. > That way, we get separation of concern (buffer vs thread). It also make more > obvious that ThreadedAuthorityFactory is buffered. Also, I'm not sure that > "ThreadedAuthorityFactory" was a better name than "BufferedAuthorityFactory" > because I believe that a majority of users will care more about buffering than > concurency (BufferedAuthorityFactory already handle many threads asking for a > same small set of CRS. ThreadAuthorityFactory handle many threads asking many > different CRS, which should be a less common case I think). > Both are trying to do the same thing; handle multiple threads. In actual fact even the direct authority factories for sql are buffered (in that they contain an optional backpointer "buffered" to their dispatcher that has a buffer). I am glad we are all understanding the design - let's get the work done. Every twitch we make on naming is costing me time, effort and trust. Jody |