Yes, completely agree. We should keep the API as practical as possible. Mike
On 12/5/05, Ortwin Glück <[EMAIL PROTECTED]> wrote: > > > Roland Weber wrote: > > That's why I rated this as esoteric. You can implement any kind of > > protocol via JNI. The protocols I can think of are either not meant > > for HTTP, or can be mapped to the Socket API. The best I can come > > up with here are pipes, but why should anybody want to use HTTP > > via pipes? > > The mail about a native implementation of SSL sockets a few days ago > > made me think of this. On second thoughts, they were still providing > > java.net.Socket, they just couldn't create the SSL socket on top of > > an existing Java socket. > > > > Back in the old days, when working on my diploma thesis and before, > > I was programming on an operating system for a message passing > > parallel computer. We didn't have anything like TCP/IP there, but > > an OS specific communication infrastructure. Of course there was no > > Java either :-) But that's why I am sometimes thinking queer and > > non-TCP, like in this case. > > > > cheers, > > Roland > > Roland, > > Thanks for explaining. But I think we should not start to implement > purely academic scenarios. It only leads to overengineered architecture > and adds complexity that is otherwise unecessary. In the end it scares > people and make the API harder to use. Unless nobody comes up and asks > if we can support HTTP over IPX or something, I wouldn't bother. > > Odi > > > --------------------------------------------------------------------- > 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]
