httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexei Kosut <ako...@nueva.pvt.k12.ca.us>
Subject Re: Last call for patches
Date Wed, 14 Feb 1996 07:35:41 GMT
On Mon, 12 Feb 1996, David Robinson wrote:

> >1. Sets r->connection->user and auth_type to NULL before each reqest.
> 
> Sigh... please move these to the request_rec structure where they belong.

Well, I would have, but they're used in a number of places besides the 
files I wanted to play with at the time, including a number of modules 
(msql and db) that are up for consideration, so this would have broken 
them. I determined that it was better just to make it work for now, and 
fix it for real next time around.

> And how about freeing up the memory used by one request before serving
> the next?

Depends. We may want to use it again at some point, by some module. 
That's why I had it store it like it did. Maybe we should make new pools 
and free them each time. Hmm.

And about Keep-Alives and proxies. *sigh* (sigh) SIGH! I just confirmed
what we already know: they don't like each other. Both CERN httpd 3.0
(prolly the most popular proxy server) and 97.proxy-03 will send the
Connection header on just fine, and then will hang on the response. CERN
will actually send it, but then won't allow the client to make another
request, so it has to time out first.

This is obviously bad. And I have no clue what to do about it. Obviously,
we should disable Keep-Alive support for these requests. But we can't. 
There is no way of telling that it is a proxy. CERN mangles the
User-Agent, but this certainly isn't common among proxies. HTTP/1.1
defines a Forwarded: header for this purpose, but no one sends it that 
I'm aware of.

Now, it has been pointed out that Netscape does not send Connection: 
Keep-Alive to proxies, and I suppose none of the other clients do, 
either. So I guess it may fall into the "hope no one notices" category. 
Except for the cases where the client doesn't know it's a proxy, of 
course. Gateways, actually, they'd be in that case.

I'd suggest our proxy module at least be modified to, as per the
suggestion in the HTTP/1.1 draft, not pass on Connection headers. That
will solve the problem with using Apache as a gateway. Especially since if
both servers supported it, you'd have a double-wait. *sigh* That would at
least solve the gateway and double-proxy problem, when using Apache's
proxy module (assuming we fix the problems I talk about in other emails). 

The HTTP/1.1 draft also says "Connection fields received in an HTTP/1.0
message, as would be the case if an older proxy mistakenly forwards the
field, cannot be trusted and must be discared except under experimental
conditions." But it appears half the web is now "experimental". *sigh*

Probably the best person to ask this of is Roy, so Roy, is that about 
right? Anything else we should/can do?

--/ Alexei Kosut <akosut@nueva.pvt.k12.ca.us> /--------/ Lefler on IRC
----------------------------/ <http://www.nueva.pvt.k12.ca.us/~akosut/>
The viewpoints expressed above are entirely false, and in no way
represent Alexei Kosut nor any other person or entity. /--------------




Mime
View raw message