On 12/6/2012 4:40 PM, Daniel Stenberg wrote:
> On Tue, 4 Dec 2012, Bill Hoffman wrote:

> I would suggest using wireshark, and once the connection has failed you
> can save just the last say 100 packets with the TCP stream and attach
> that pcap file in a follow up here. Could reveal something.

I have attached the pcap file. I sent a larger file to Daniel off the
list and he pointed out the interesting bits. I have attached the
smaller interesting part of the data.
>
>> One more small piece of information. We took a successful download of
>> a file and chopped off the end of it to make it match the size of a
>> failed file and they were still different. So, it seemed that there
>> was something missing from the middle of the file.
>
> That is indeed very curious. Can you investigate how they differ and if
> there's a pattern over repeated test runs (and failures) ?
>
  One thought that might explain the server sending an early FIN would
be if somehow the client missed something in the middle. If a packet
was dropped, and the server thought it had sent it, but the client did
not get it, that would cause an early FIN. It would explain why the
files are different earlier on.

I will see if I can find a pattern.

-Bill

-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html