I have been experiencing problems with the way passive-close is specified for DCCP and developed a solution which avoids the following problem: When a side receives a connection termination (Close or CloseReq), it can happen that the receive buffers are wiped out before the application had a chance to see the data; sometimes even before connect() returns. The solution is based on implementing transitional states (PASSIVE_1/2) for passive-close. As in TCP, these are terminated after a timeout. I have put the concept and blurb online: http://www.erg.abdn.ac.uk/users/gerrit/dccp/docs/closing_states/ There are patches for this which I would like to submit along with the remaining ones for CCID3. I'd be glad if people could have a look at this and say if there are disagreements with the concept. Note: This implementation concept does not affect standards-compliance since the macroscopic behaviour is still compatible with RFC 4340. I am however quite positive that without transitional states it will not work and the API will only function for one or two special scenarious of socket programming. Gerrit - To unsubscribe from this list: send the line "unsubscribe dccp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
