Keep-Alive and Content-Length problem

Hi!

I hope this is the right newsgroup for this problem - I have searched all
around the web and couldn't find the answer.

Short summary:
Apache + PHP on some servers return "Connection: Keep-Alive" header, but
without "Content-length: ..." header. Is this correct? How does client know
when the content ends?

Background:
I have a CMS which is (of course) locked by authentication. I'm using
PHPlib, which in turn uses PHP sessions. The whole system is well tested
over last 5 years or so and always worked flawlessly.
Lately there has been an occasional problem logging into the application.
The problem is server specific, but what is even weirder, it is also
browser-specific. For instance, I can always login without problems in
Konqueror, usually without problems in IE and... Firefox makes trouble.

Because of this I installed FF extension livehttpheaders and listened to
traffic between browser and server and compared it to a server where
everything worked. The only difference was Keep-Alive header:

GET /page.php HTTP/1.1
Host: www.domain.com
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.2)
Gecko/20070208 Iceweasel/2.0.0.2 (Debian-2.0.0.2+dfsg-3)
Accept: text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,
text/plain;q=0.8,image/png,*/*;q=0.5
Accept-Language: en-us,en;q=0.5
Accept-Encoding: gzip,deflate
Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
Keep-Alive: 300
Connection: keep-alive

HTTP/1.x 200 OK
Date: Mon, 19 Mar 2007 12:32:56 GMT
Server: Apache
Set-Cookie: PHPSESSID=bab0ea73140b093fa952396f22075a9c; path=/
Expires: Thu, 19 Nov 1981 08:52:00 GMT
Cache-Control: no-store, no-cache, must-revalidate, post-check=0,
pre-check=0
Pragma: no-cache
Keep-Alive: timeout=3, max=600
Connection: Keep-Alive
Transfer-Encoding: chunked
Content-Type: text/html; charset=ISO-8859-2

Then the client starts the next connection (for some image) and so on.

Now: there is Keep-Alive header, but no Content-Length. How does the browser
know when the content is finished?

Hope it makes sense... :)

Best,

Anze

On 19 Mar, 13:05, Anze <anzen...@volja.net> wrote:
> Hi!
>
> I hope this is the right newsgroup for this problem - I have searched all
> around the web and couldn't find the answer.
>
> Short summary:
> Apache + PHP on some servers return "Connection: Keep-Alive" header, but
> without "Content-length: ..." header. Is this correct? How does client know
> when the content ends?
>
> Background:
> I have a CMS which is (of course) locked by authentication. I'm using
> PHPlib, which in turn uses PHP sessions. The whole system is well tested
> over last 5 years or so and always worked flawlessly.
> Lately there has been an occasional problem logging into the application.
> The problem is server specific, but what is even weirder, it is also
> browser-specific. For instance, I can always login without problems in
> Konqueror, usually without problems in IE and... Firefox makes trouble.
>
> Because of this I installed FF extension livehttpheaders and listened to
> traffic between browser and server and compared it to a server where
> everything worked. The only difference was Keep-Alive header:
>
> GET /page.php HTTP/1.1
> Host:www.domain.com
> User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.2)
> Gecko/20070208 Iceweasel/2.0.0.2 (Debian-2.0.0.2+dfsg-3)
> Accept: text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,
> text/plain;q=0.8,image/png,*/*;q=0.5
> Accept-Language: en-us,en;q=0.5
> Accept-Encoding: gzip,deflate
> Accept-Charset: ISO-8859-1,utf-8;q=0.7,*;q=0.7
> Keep-Alive: 300
> Connection: keep-alive
>
> HTTP/1.x 200 OK
> Date: Mon, 19 Mar 2007 12:32:56 GMT
> Server: Apache
> Set-Cookie: PHPSESSID=bab0ea73140b093fa952396f22075a9c; path=/
> Expires: Thu, 19 Nov 1981 08:52:00 GMT
> Cache-Control: no-store, no-cache, must-revalidate, post-check=0,
> pre-check=0
> Pragma: no-cache
> Keep-Alive: timeout=3, max=600
> Connection: Keep-Alive
> Transfer-Encoding: chunked
> Content-Type: text/html; charset=ISO-8859-2
>
> Then the client starts the next connection (for some image) and so on.
>
> Now: there is Keep-Alive header, but no Content-Length. How does the browser
> know when the content is finished?
>
> Hope it makes sense... :)
>
> Best,
>
> Anze

does it have to know?
mostly they take a stream and render it if they can.
or use the javascript when instructed, even if corrupt, css etc...

keep live headers on and browse round the net, you will see very many
cases where the cl header is not set.

as for the content length header in your application, this is merely a
symptom of config, either in your server, or more probably your app,
so I'm suggesting that the header isn't the only difference.

Why not hack your app to send that header, test it, and if it works,
you've solved your issue. Personally I can't see /if the content-
length header _is_ the only difference/ that it would make a
significant difference to the functionality you're talking about - it
might though - I just can't see how at the mo.

Can you tell us more config diff between this and the working box(es)?

Anze wrote:

Your question:

> Apache + PHP on some servers return "Connection: Keep-Alive" header, but
> without "Content-length: ..." header. Is this correct? How does client
> know when the content ends?
>

Your answer:

> Transfer-Encoding: chunked

Try Google for the specifics of chunked encoding.

C.

Thank you both for your answers!

Yes, chunked encoding is the answer... We live to learn. :)
Well, I'll just have to debug more.

Thank you again for clearing this up and happy coding!

Best,

Anze


> Your answer:
>
>> Transfer-Encoding: chunked
>
> Try Google for the specifics of chunked encoding.

>> Now: there is Keep-Alive header, but no Content-Length. How does the
>> browser know when the content is finished?
>
> does it have to know?
> mostly they take a stream and render it if they can.

Yes, they have to know - otherwise page rendering wouldn't be much stable.
Of course they try to render a page asap, but still... Coding a browser
that wouldn't know when a page ended would be unnecessarily much harder.

But this is solved through chunked encoding which I wasn't aware it existed.

Thank you for your answer!

Best,

Anze