On Wed, 2007-11-14 at 13:38 +0100, Ortwin Glück wrote: > > Oleg Kalnichevski wrote: > > (2) We do have some home-baked concurrency primitives in HttpCore NIO, > > but none of them is exposed in the public API. The potential reduction > > of custom code does not seem significant enough to justify the change of > > minimal JRE level just a couple of weeks before the API freeze. > > Synchronization that can be avoided with stuff from the concurrent > package? (Not because of performance, but because of the danger of > unforseen deadlocks.) >
The I/O reactor code is fairly simple from the synchronization standpoint. Really complex stuff usually happens at the protocol handling level and is supposed to be outside Core anyways. Someone (preferably not me, because I can be biased) should take a fresh look at the HttpCore NIO and decide whether the use of j.u.concurrent would result in significant code improvements. Cheers Oleg > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
