Bug 53590

Description of Problem:

When recursively descending large web sites (mirroring), wget will 
sometimes get into a state where it seems to get stuck in a loop (I 
usually use -nc so this consists of "xxx already there, will not 
retrieve, yyy already there will not retrieve, xxx already...").

I can't provide a meaningful test case that is of usable size. I'm going 
to go on a spelunking mission in the source to see if I can track it down 
using just my understanding of the source code (I've patched this sucker 
a number of times, so I'm already familiar).

Right now, suggest you either (1) wait a few days until I either come up 
with something or concede concede defeat or (2) try the same course.

TODO: see the most recent issue of MSDN. There's an article about a 
program for IIS (which would translate to a simple module or CGI for a 
real web server) that uses statistical analysis to kick mirroring 
programs off the site. I'm going to patch wget to select a random number 
from a range of "time-to-wait-between-connect" values to defeat it. Also, 
note the irresponsible and brain-dead suggestion that it would be best to 
only use the class C net address of the client to identify visitors, 
since DHCP defeats it otherwise. E.g.: I have a 4 bit netblock on a 
registered domain (geeksrus.net). Someone else I've never heard of on the 
same class C gets me banned from a website. Or you're on DHCP, so one guy 
bans 252 other people. Nice.

server)



Version-Release number of selected component (if applicable):


How Reproducible:


Steps to Reproduce:
1. 
2. 
3. 

Actual Results:


Expected Results:


Additional Information: