On Sat, 9 Oct 1999, Alan Cox wrote:
@>
@>> 17:52:08.920000 THEM.www > ME.2900: . 88202:89174(972) ack 625 win 9720
<nop,nop,timestamp 37623391 243244922> (DF)
@>>
@>> 17:52:09.260000 THEM.www > ME.2900: . 89174:90146(972) ack 625 win 9720
<nop,nop,timestamp 37623391 243244922> (DF)
@>>
@>> 17:52:09.260000 ME.2900 > THEM.www: . ack 90146 win 31104 <nop,nop,timestamp
243246035 37623391> (DF)
@>
@>Note the ack value here is 90146
@>
@>> 17:52:09.620000 THEM.www > ME.2900: . 92090:93062(972) ack 625 win 9720
<nop,nop,timestamp 37623529 243245061> (DF)
@>
@>This data is after a frame is missed.
@>
@>> 17:52:09.620000 ME.2900 > THEM.www: . ack 90146 win 31104 <nop,nop,timestamp
243246071 37623529,nop,nop,sack 26369@63727 27341@63727> (DF)
@>>
@>> 17:52:09.960000 THEM.www > ME.2900: . 93062:94034(972) ack 625 win 9720
<nop,nop,timestamp 37623529 243245061> (DF)
@>
@>We keep acking the old value, and it keeps moving on sending newer and newer
@>valid data but always with the hole. It looks like you may have a modem config
@>problem or something similar that stops some specific bitpattern from being
@>passed.
@>
@>
Thankyou for your response;
If what you guess were true, wouldnt the SACK option grow to eventually
include 3 ranges around the holes?
I think the appearance of THEM sending increasing sequence numbers and
appearing to send new data is a bug or feature when tcpdump is not using
the real sequence numbers, but their offset. Here is some of the final
output from another attempt to download, but with real sequence numbers
and the SACK option displayed correctly(?):
20:32:02.650000 THEM.www > ME.4.3170: . 3903557709:3903558681(972) ack 3908018072 win
9720 <nop,nop,timestamp 38583497 244178384> (DF)
20:32:02.650000 ME.3170 > THEM.www: . ack 3903558681 win 30824 <nop,nop,timestamp
244205374 38583497,nop,nop,sack 3903559373 3903588533> (DF
)
...(repeats forever)...
If I am interpreting the SACK value correctly, then our hole is in the
window beyond the one they keep resending. This would mean the other
side isnt behaving right?
Im uncomfortable believing that SunOS 5.7 would do this (THEM)... How
commonly used is SACK?
Unfortunately, my connection seems to be good enough today, that I cant
get it to happen again...
Paul
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
