On Dec 28, 2004, at 6:03 AM, Alex Tweedly wrote:
When you 'open datagram socket', specify a callback. The callback is NOT for completion of the open. That completes immediately. It is for the incoming packets, the responses.
I could have tried for a month of Sundays without guessing that - thank you.
Me, too. I learned about from Scott Raney when I reported a bug. Somehow, what I had been doing was working.
I think there is a motivation for TCP to do it the UDP way. I wonder if there is a hole in the Revolution model for TCP in which a data packet with a FIN can come in during the gap between reads and the data get lost.
I've been working with connectionless protocols including UDP for the best part of 20 years; I'm pretty comfortable with distributed algorithms - the model fits this application very nicely.
Me, too. We might be resources for each other.
It's simply the translation of the standard BSD-socket API into Rev that's confusing me (and the rather less than helpful docs).
I don't mind the Rev API. Doing away with select is a great plus, the weakest link in BSD sockets. I do find things I want to do such that I can't. You mentioned getting the local port. I have a protocol in which I need to set it.
I had proposed a peer-to-peer model for Revolution UDP some time back, but when it was set up, it was set up as client-server and used accept. (One problem with that is that if you set up a server for both UDP and TCP on the same port, then they look the same on openSockets() and to close socket.)
Dar
**********************************************
DSC (Dar Scott Consulting & Dar's Lab)
http://www.swcp.com/dsc/
Programming Services and Software
**********************************************_______________________________________________ use-revolution mailing list [email protected] http://lists.runrev.com/mailman/listinfo/use-revolution
