On Fri, 2007-07-20 at 18:28 +0200, David J. M. Karlsen wrote: > Oleg Kalnichevski wrote: > > HttpClient 4.0 represents a complete, ground-up redesign and almost a > > complete rewrite of the HttpClient 3.x codeline. This release finally > > addresses several design flaws that existed since the 1.0 release and > > could not be fixed without a major code overhaul and breaking API > > compatibility. > > > > The HttpClient 4.0 API is still very experimental and is bound to change > > in the course of the ALPHA development phase. Several important features > > have not yet been ported to the new API. > > > Cool - good work. > Just curious (not critisism) - why not base this on the Apache MINA project? >
David, (1) There are good and bad things about MINA. While it has a nice and elegant API I _personally_ consider MINA's way of doing certain things pretty sub-optimal. For instance, maybe for some protocols creating a Future object for _every_ write operation makes sense, but I personally fail to see a point in doing so for HTTP. I also tend to dislike the way MINA manages memory. Again, this is just my personal, very humble opinion. (2) NIO tends to perform significantly worse than the classic blocking I/O in terms of data throughput as long as the number of worker threads is less then a thousand. I do not see a reason why a pure HTTP client may need over a thousand worker threads. Oleg --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
