Hi, On Fri, 4 Jul 2003, fish wrote:
> FCP Bulk transfers are giving me incredible amounts of grief in both stable and > unstable right now, and I'm clueless as to why. So am I. Started to develop my own FCP-Libabry a while ago, for the mere testing. But got stuck with the same problems you are facing. > The short version is, if I throw data at the node too quickly, the connection > will hang. "Too Quickly" seems to be incredibly arbitary, sometimes i can > feed the beats at 1kbytes/sec, someitmes it will even fail at 512bytes/sec, > someytimes it will succeed at as high as 10-20kbytes/sec. The chance of > success seems to increase the slower that I feed it. My transfers usually fail when about 70-90kB are already transferred. I have not experimented with throttling, for I thought the node should not accept data if it was not able to process it fast enough. > This is complete alchemy, and I have no understanding of what it occours, > or if I'm just triggering something else completly unrelated and getting > lucky and clinging onto anything which looks like an explanation that > I can understand :) I have not experienced this with ClientPuts, only with FECEncodeSegment Commands so far. First I thought it was because I ommited the MetadataLength field in the Command, which is by the way not in the Documentation, Frost nevertheless uses it. But it does not make it more functional. Some FEC Commands in FCP appear not to issue FormatErrors at all, so you can never be sure if the Node understood the command. > I've heard that FIW works, so I assume that I'm doing something > differenet to everyone else - does anyone have any idea what could > cause this behaviour? No, unfortunately not. But it sounds very familiar to me ;( Stefan _______________________________________________ devl mailing list [EMAIL PROTECTED] http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
