"Can't read stream because the TCP connection..."

Doc Searls Manila-Newbies@userland.com
Tue, 9 Jul 2002 13:06:14 -0700

At 10:25 AM -0700 7/9/02, Sam DeVore wrote:
>On 7/9/02 12:35 AM, "Doc Searls" <doc@searls.com> wrote:
>>  At 4:17 PM -0700 7/8/02, Sam DeVore wrote:
>>>  Sounds like the same bug in the networking that bites me when trying to
>>>  upstream to a rcs host running windows.  Let me guess Doc, running os x?
>>>  Did you change anything about your network recently?
>Doc on the ping test what kind of failure rate are you getting?
>Network utility will report this when done. Andre, Doug and I did a bunch of
>testing with this (I suffer from the same problem) and it seemed to come
>down to the fact that os x connecting to a windows machine do not handle
>failed packets right.  The problem doesn't seem to happen when os x connects
>to os x or os x to os 9.  There is some issue with the connection to
>windows, I am beginning to wonder if it related to the the recent
>revelations about windows not running at full duplex when talking to os x
>that I have been reading about.

Here's the whole dialog in Network Utility:

PING ( 56 data bytes
64 bytes from icmp_seq=0 ttl=237 time=95.621 ms
?Q64 bytes from icmp_seq=1 ttl=237 time=94.274 ms
64 bytes from icmp_seq=2 ttl=237 time=93.371 ms
64 bytes from icmp_seq=3 ttl=237 time=94.923 ms
64 bytes from icmp_seq=4 ttl=237 time=94.207 ms
64 bytes from icmp_seq=5 ttl=237 time=96.301 ms
64 bytes from icmp_seq=6 ttl=237 time=97.55 ms
64 bytes from icmp_seq=7 ttl=237 time=94.108 ms
64 bytes from icmp_seq=8 ttl=237 time=94.36 ms
64 bytes from icmp_seq=9 ttl=237 time=94.756 ms
64 bytes from icmp_seq=10 ttl=237 time=93.97 ms
64 bytes from icmp_seq=11 ttl=237 time=93.647 ms
64 bytes from icmp_seq=12 ttl=237 time=94.878 ms
64 bytes from icmp_seq=13 ttl=237 time=94.985 ms
64 bytes from icmp_seq=14 ttl=237 time=93.788 ms
64 bytes from icmp_seq=15 ttl=237 time=97.868 ms
64 bytes from icmp_seq=16 ttl=237 time=94.443 ms
64 bytes from icmp_seq=17 ttl=237 time=93.467 ms
64 bytes from icmp_seq=18 ttl=237 time=95.008 ms
64 bytes from icmp_seq=19 ttl=237 time=94.296 ms
64 bytes from icmp_seq=20 ttl=237 time=96.422 ms
64 bytes from icmp_seq=21 ttl=237 time=95.012 ms
64 bytes from icmp_seq=22 ttl=237 time=93.575 ms
64 bytes from icmp_seq=23 ttl=237 time=94.973 ms
64 bytes from icmp_seq=24 ttl=237 time=107.706 ms
64 bytes from icmp_seq=25 ttl=237 time=94.375 ms
64 bytes from icmp_seq=26 ttl=237 time=93.223 ms
64 bytes from icmp_seq=27 ttl=237 time=111.273 ms
64 bytes from icmp_seq=28 ttl=237 time=110.23 ms
64 bytes from icmp_seq=29 ttl=237 time=111.322 ms

--- ping statistics ---
30 packets transmitted, 30 packets received, 0% packet loss
round-trip min/avg/max = 93.223/96.797/111.322 ms

The IP address is Yahoo's, fwiw. BSD.

The problem is slow upstream speed. I'm getting between 2 and 44Kb, 
even though the downstream side is around 3Mb. Mostly it's around 2. 
Attempts to FTP upstream just sit there. The blog routine fails. 
About 1 in 5 upstream SMTP semail attempts fail, though POP3 calls go 
through fine. Taking out hubs and routers make no difference. I've 
tested it with Windows (via VPC), OS 9 and OS X on different 
machines, both on the router and directly to the cable modem, on a 
variety of cables. Same thing every time.

They're sending a guy out between 12 and 2pm tomorrow, since I 
finally convinced them there's something fucky on the Ethernet side 
of the cable modem. On the cable side it tests at 300kb upstream and 
3.xMb downstream. On the ethernet side, it's fine downsteam and 
terrible up.

Very frustrating.