https://issues.apache.org/bugzilla/show_bug.cgi?id=49391
--- Comment #3 from Graham Leggett <[email protected]> 2010-07-01 20:52:46 EDT --- It's even more complex still. We need cache updates to be atomic, right now they are not. The idea of having a well-known filename for the header, and a well-known filename for the data is currently bogus, as you can't update both at the same time. What we need to do is give a unique name to the data, and then store that unique name within the headers file. We are then in a position to write the data body in place at leisure, and when done, we insert that data filename into the well-known header file, and write the header file atomically. If the connection is aborted for whatever reason before the header is written, we simply blow away the data file in the cleanup, and we're done. After the header is written the cache's work is done, so there is no need for a cleanup. The order that the hooks are called is not important from mod_disk_cache's perspective, mod_disk_cache is not obligated to write the headers when it receives them. Instead it can set aside the headers and delay the writing of the headers to the point at which we receive the final data and EOS is detected. -- Configure bugmail: https://issues.apache.org/bugzilla/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
