I tried HTTrack, it is a great tool to mirror a web site. 
Now I want to use HTTrack to setup a test web server that 
keeps instances of some URLs, the requirement is when the 
tester visits a test URL (i.e. www.msn.com) through test 
web server, this test web server should provide all 
components (www.msn.com/index.html and all embedded 
components associated with www.msn.com/index.html) just 
like he visits same URL through original web server. It is 
not necessary to download related links (i.e. Autos 
Careers) because we are only interested with the 
performance to download this page. The reason to setup 
this test bed is to get controlled environment to test 
compression servers.  Because the contents in test web 
server are static and the connection between compression 
servers and test server is perfect, we can make sure the 
difference of test results is only contributed by 
different compression servers instead of changed contents 
and network environment. 

I tried using HTTrack -w URL, it works fine for some sites 
like www.netzero.net, but it does not work well for some 
popular sites like sports.yahoo.com, www.ebay.com and 
www.amazon.com. For some sites, contents of test web 
server downloaded by HTTrack can't provides all required 
components (I used TCPDUMP to monitor all download 
traffic) therefore the browser has to visit original server
(s) to download some embedded components that violates the 
design of the test bed.

For www.msn.com <http://www.msn.com>, it seems that I can 
download all components but the appearance in browser when 
visiting through the test server is different from when 
visiting through original www.msn.com <http://www.msn.com> 
server that means the browser may get different objects 
between the test server and original web server.

I run HTTrack on Linux 2.4.24; the web server is Apache 
2.0.40 . Client machine is Windows XP home edition with IE 
6.0.
  

Could you please give me some suggestions to use your 
HTTrack to setup the test bed?
Thanks

Fan