httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From dean gaudet <dgaudet-list-new-ht...@arctic.org>
Subject Re: woah, "GET /" with autoindex
Date Sun, 07 Jan 2001 19:57:41 GMT
On Sun, 7 Jan 2001, dean gaudet wrote:

> woah.  ok there's some serious problems here.
>
> this is from a "GET / HTTP/1.0" which results in an autoindex (dunno why
> it's not grabbing one of the index.htmls, but i'm not worrying about that,
> the autoindex is probably far more interesting).

btw, the tcpdump for this shows suboptimal packets as well.  this is
because without nagle, we've told the kernel that each of our writev()s is
to be put on the network.

11:53:12.839439 eth0 < client.1131 > server.8888: S 1366931848:1366931848(0) win 32120
<mss 1460,sackOK,timestamp 983960148 0,nop,wscale 0> (DF)
11:53:12.839531 eth0 > server.8888 > client.1131: S 1369163235:1369163235(0) ack 1366931849
win 32120 <mss 1460,sackOK,timestamp 769294182 983960148,nop,wscale 0> (DF)
11:53:12.856197 eth0 < client.1131 > server.8888: . 1:1(0) ack 1 win 32120 <nop,nop,timestamp
983960150 769294182> (DF)
11:53:14.969574 eth0 < client.1131 > server.8888: P 1:17(16) ack 1 win 32120 <nop,nop,timestamp
983960361 769294182> (DF)
11:53:14.969626 eth0 > server.8888 > client.1131: . 1:1(0) ack 17 win 32120 <nop,nop,timestamp
769294395 983960361> (DF)
11:53:15.109544 eth0 < client.1131 > server.8888: P 17:19(2) ack 1 win 32120 <nop,nop,timestamp
983960375 769294395> (DF)
11:53:15.119937 eth0 > server.8888 > client.1131: . 1:1(0) ack 19 win 32120 <nop,nop,timestamp
769294411 983960375> (DF)
11:53:15.139046 eth0 > server.8888 > client.1131: P 1:449(448) ack 19 win 32120 <nop,nop,timestamp
769294412 983960375> (DF)
11:53:15.139164 eth0 > server.8888 > client.1131: P 449:541(92) ack 19 win 32120 <nop,nop,timestamp
769294412 983960375> (DF)
11:53:15.167745 eth0 < client.1131 > server.8888: . 19:19(0) ack 449 win 32120 <nop,nop,timestamp
983960381 769294412> (DF)
11:53:15.167806 eth0 > server.8888 > client.1131: P 541:1989(1448) ack 19 win 32120
<nop,nop,timestamp 769294415 983960381> (DF)
11:53:15.167821 eth0 > server.8888 > client.1131: P 1989:3437(1448) ack 19 win 32120
<nop,nop,timestamp 769294415 983960381> (DF)
11:53:15.174817 eth0 < client.1131 > server.8888: . 19:19(0) ack 541 win 32120 <nop,nop,timestamp
983960382 769294412> (DF)
11:53:15.174882 eth0 > server.8888 > client.1131: P 3437:4885(1448) ack 19 win 32120
<nop,nop,timestamp 769294416 983960382> (DF)
11:53:15.174904 eth0 > server.8888 > client.1131: FP 4885:5072(187) ack 19 win 32120
<nop,nop,timestamp 769294416 983960382> (DF)
11:53:15.207674 eth0 < client.1131 > server.8888: . 19:19(0) ack 3437 win 31856 <nop,nop,timestamp
983960385 769294415> (DF)
11:53:15.227760 eth0 < client.1131 > server.8888: . 19:19(0) ack 5073 win 30408 <nop,nop,timestamp
983960387 769294416> (DF)
11:53:15.238239 eth0 < client.1131 > server.8888: F 19:19(0) ack 5073 win 31856 <nop,nop,timestamp
983960388 769294416> (DF)
11:53:15.238296 eth0 > server.8888 > client.1131: . 5073:5073(0) ack 20 win 32120 <nop,nop,timestamp
769294422 983960388> (DF)

1.3 would generate three 1448 byte packets plus one straggler.

on linux you could CORK in 2.0 and unCORK at the end and it'd generate the
right packets... but that's not portable.  so the fix needs to live
somewhere in the i/o layering... which i'm not familiar enough with yet.

-dean


Mime
View raw message