Kevin Smith;1753 Wrote: > On Tue, Jul 15, 2008 at 4:35 PM, Jehan > <Jehan.3clj1z (AT) no-mx (DOT) jabberforum.org> wrote: > > Anyway don't you think this would be better to improve client and > > server implementations rather than adding a new layer atop all the > > current one? XMPP is made to be exchanged on top of a reliable > > connection (most common and the one it has mostly been designed for > > being TCP!). In such context you already have aknowledgments about > > succeeded transfer or not. > > Oh if it were that simple :) > I believe this has been discussed before at greater length than I have > the knowledge to repeat but the summary is: > 1) Your application may not be able to get the TCP/IP state directly > enough from the stack to know about failures. > 2) Your TCP/IP stack, or your provider's, may be neutered such that > the ACKs don't mean anything useful anyway. > 3) You might be using a different transport. > > /K
Ok, if you say so. I know the TCP protocol theoritically from university but it is true I have never been interested enough to look at the real implementations. So all I really know about TCP (else than theory) is in fact the different available APIs of socket programming. Then I guess a stanza aknowledgement is useful. But just for my own information: do you know the opinion of network developpers about this flaw when losing a connection (which is common on unreliable network materials nowaydays)? Do they just consider this as normal that their implementation does not report to the higher layers a failure of data transfer because of unexpected closed connection? Or are developpers trying to fix this? Jehan -- Jehan ------------------------------------------------------------------------ Jehan's Profile: http://www.jabberforum.org/member.php?userid=16911 View this thread: http://www.jabberforum.org/showthread.php?t=417
