Justin Mason writes, > If completewhois is still down, and has been having trouble since > people started making comments about it on the users list, we > should probably remove it. > http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5632
I don't mind removing it, although I have no idea how long it has been down and what is their status. If it's just been under attack, removing it would better be a temporary measure only, perhaps through channels. Btw, looking at a timing report from these failing queries revealed an optimization opportunity, which can save up to one second of elapsed time (if HiRes::Time is available), most pronounced under just such conditions where some RBL is timing out: Sending lib/Mail/SpamAssassin/AsyncLoop.pm Transmitting file data . Committed revision 570127. Shave off up to 1 second of wait time during the final call of complete_lookups() from harvest_dnsbl_queries(), if all queries have expired by the current time + 1 second. Explanation: harvest_dnsbl_queries polls in full 1 second intervals. Even if all queries expire at some time during the last 1 s interval, the 'select' would still wait for the full 1 second period. Now the last 1 s wait interval is reduced to only cover up to the latest query expiration time. This optimization is made possible by the subsecond timing. Mark
