> Aaron Stone wrote: > >> Hmm, yeah, true, no decision yet, but lots of excitement ;-) However, >> there are some things that simply need to be threaded. Afaict, Oracle's >> connection overhead is so huge that nothing getting near it can survive >> without threading. I'd like to see some of that with my own eyes, but >> I'm >> confident that the numbers I have read over the years aren't entirely >> off >> base. >> >> >> > Could you please explain how Oracle's connection overhead directly > translates to we need threads? A new thread still has to open up a new > connection to the DB. It sounds to me that Oracle's connection overhead > (or Postgresql or whatever) is more an argument for persistent > connections and pre forking more than threads. But if I'm missing > something please let me/us know.
I know a good bit about oracle (and maybe re-writing a new driver for it) but that wasn't my driving force behind the code work... as for my implementation it uses thread pool's and so it only opens a connection the first time a thread is used then it keeps that connection open in memory for the next client and then closes all of them (up to MAXCHILDREN) at the end of the life of the daemon. It works really nicely so far for the pop3d, I wan't to pose a question about the message cache code in the imapd which is giving me some serious issues, any reason dbmail shouldn't convert to using GCache (glib) and think about using GMime for message parsing, http://spruce.sourceforge.net/gmime/. GMime in my research last night ( and early this morning) would allow us easy access to all the header variables with resonable high performance to do the header caching code in the delivery and then allow for re-parsing for the imap code. I am going to be tring it maybe today in my quest for a thread safe imapd, it is not easy :/ Thanks, Leif p.s. I don't expect any of my patch to be accepted however I belive that I am doing a resonably good job and would be glad if parts of it were considered to be incuded. For all in dbmail land you know I have been lurking on dbmail-dev ...etc for a long while and have only contributed a little up till now, I have a reson for all this code work as I have a contract pending a proof of concept for dbmail that would put dbmail on a 200+ node cluster (w/ mericom..etc.) and do some serious mail, obviously even with pthreads it won't run on the cluster with any reasonable usablity but the idea is to look into the thread safeness of the code and then be able to extend it into a modular system...etc all in concept right now but I thought it would be nice to at least give the proof of concept code back to dbmail and see what you all will use/like from it.... havn't had much sleep sorry rambling...
