🚀 go-pugleaf

On Thursday, in article <dFMTwNj030n@sktb.demon.co.uk>
     pla@sktb.demon.co.uk "Paul L. Allen" wrote:

> In article <876390153snz@isis.demon.co.uk>
>     ian@isis.demon.co.uk (Ian Smith) writes:
>  
> > Can anybody enlighten me about the command format mainstream 
> > browsers use to call a proxy server ?
> 
> The details of HTTP 1.1 are in RFC 2068 and HTTP 1.0 in RFC 1945.  I
> don't think HTTP 0.9 ever made it out of an IETF draft although it was
> used for a long time.  Servers tend to follow what the specs say more
> closely than browsers do, though servers also accepted verious pieces
> of crap sent by broken browsers (like Netscape and IE).  If you do what
> the spec says, a server will accept it.  Well, with one exception.  Most
> servers still do not accept HTTP 0.9 requests although both 1.0 and 1.1
> say this is a mandatory requirement.

I did look at 1.1, but it scared the life out of me. 1.0 looked much 
cruder and so a lot more my style :) 1.1 is for next year...

> > I'm (reluctantly) putting a proxy server option into John Washington's 
> > KA9Q HTTP client and the command format I'm testing is:
> > 
> > GET proxy_server:port full_URL HTTP/1.0
> 
> Ummmm, you're not sending that, are you?  It should be GET full_URL HTTP/1.0.

No, you're right. I've awarded myself a smack 'ead !  I'd done the code 
a while ago and just returned to it. I should have reread what I'd 
actually written and not gone from memory :(

What actually happens is that I log into proxy_server:port and send

      GET full_URL HTTP/1.0

> > which appears to work and so far as I can see is all that's required 
> > with HTTP/1.0 (the HTTP/1.0 RFC is a bit spare with the proxy info),
> 
> Well, it's all that is required but you should strongly consider sending
> 
>   GET full_URL HTTP/1.0
>   Host: fqdn_extracted_from_URL
>   <CR>
> 
> Note that you need a blank line after HTTP 1.0 or greater requests to
> indicate that all headers that you want to supply have been supplied (the
> 1.0 spec didn't make that clear enough, can't remember if the 1.1 spec
> does).  Otherwise it's going to break when Demon switch the homepages
> service to relying on the Host header being present rather than using a
> load of static IP addresses.

I've been sending the Host: line for a while since it's one of the 
things required these days if you not to be refused with a snotty 
message that your browser isn't up to it, which is probably what you're 
alluding to. I've also got a user agent line (which can be spoofed) as 
another line of attack on the refusal thing although the UA tends to 
change some of what you get rather than fix a refusal. However, it also 
registers as 0.00% browser usage on Browserwatch :)

> You also need to support the POST method for transmitting long form data.

Hmmm, I have looked at that, but it looks messy and I've yet to find a 
need great enough for the effort. Almost the only things I ever send are 
search terms to search engines and they don't need POST. I think 
if-modified-since will come before POST.

> > but  when I did a trace on what Netscape 2.0 sent it didn't appear to
> > include the proxy name
> 
> Good.  Where did you get the idea that it should?

From the murky and unreliable depths of my memory, I'm afraid :(

Ian.
-- 

  |  Ian Smith             |  "The Moving Finger writes;
  |  ian@isis.demon.co.uk  |   and, having writ, Moves on."