httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Behlendorf <>
Subject Re: Problem with non-IP virtual hosts - BAD interaction with web cache servers! (fwd)
Date Sun, 24 Nov 1996 20:12:28 GMT

> From: WWW server manager <>
> I recently provided advice to a local site running Apache on how to set up
> "non-IP" virtual hosts on an existing server, including the advice on how
> to set up links in order that both old (e.g. lynx) and new (e.g. Netscape 2/3)
> users would see consistent, usable links.
> Access seems to be working as expected except for one problem, which (if my 
> diagnosis is correct) will be a more general problem and will need more 
> specific advice to avoid confusing results for some users, in some 
> situations.


> Does this analysis sound plausible? Do you have any alternative ideas for how 
> to get around the problem, either now (Apache 1.1.1) or by changes in future 
> versions?

Thanks for that thorough explanation of interesting cache interactions with
non-IP-intensive vhosts.  I'd actually like to save that and put it on the site at some point, if you don't mind, or somehow integrate it
into a discussion of Host:-header-based servers.  

I know there was a solution discussed on the HTTP IETF working group mailing
list on this subject, though I can't seem to find it right now (after 1/2 hour
of searching).  I think some people feel this is just a wart with which we live
until proxies get upgraded to HTTP 1.1, and start adding Host: headers to
requests which don't have them.  The other option would be to hack up apache to
always treat requests without Host: headers as non-cacheable - i.e. send a
Pragma: no-cache.  That at least would deal with one of your two scenarios,
that of the old lynx user loading up the cache with "incorrect" pages, but not
the second case, where the netscape 3 user went first followed by the lynx
user.  Lynx 2.5 and later, by the way, do implement the Host: header; and I
don't have exact statistics as to what percentage of requests to a typical
server are from Host-less browsers, but it is decreasing rapidly.  

I guess I don't know what the answer could be.  I know that I'm not a fan of
using full-URL links off of html pages, as you suggested, because even as you
state it's not going to fix the problem, and it is going to be a tremendous
headache for folks, unless you're suggesting the server rewrite it as the pages
are served or something.

Anyways, thanks again for the analysis.  Some other folks here on the apache
list may have more comments.



View raw message