ID: 25175
User updated by: phpbug at paypc dot com
Reported By: phpbug at paypc dot com
Status: Open
Bug Type: Output Control
Operating System: Linux 2.4
PHP Version: 4.3.2
New Comment:
One last thing... this bug report isn't just a pedantic exercise.
When I have that defective Content-Length header in the headers, IE
5.00.3700.1000 SP4 under Windows 2000 Profession SP4 "hangs" for a few
seconds during a pageview when that header is present. When I comment
out the Content-Length header invocation, this problem goes away. NO
INTERVENINIG PROXIES ARE INVOLVED - as it's a local intranet without
even an intevening router in between the two machines.
I am to understand that HTTP pipelining and keep-alive support is
dramatically assisted by the presence of this header. I would like to
make use of it.
I am going to sprinkle my preamble+postamble across a wider mixture of
PHP pages on my system (some with DB content and some without) to see
if I can find common modes of failure.
If any Zend/PHP devs can suggest a more surgical strike, I'd sincerely
appreciate it.
=Apollyon=
Previous Comments:
------------------------------------------------------------------------
[2003-08-20 20:45:00] phpbug at paypc dot com
Furthermore, when I removed the Keep-Alive header, nothing changes.
IN FACT, my preamble file is now:
header.inc:
------------------------------
<?php
$start_time=gettimeofday();
ob_start();
?>
------------------------------
footer.inc:
------------------------------
<?php
header('Content-Length: ' . ob_get_length());
ob_end_flush();
?>
---------------------------------
The headers output from a page:
HTTP/1.1 200 OK
Date: Thu, 21 Aug 2003 01:36:53 GMT
Server: Apache/1.3.27 Ben-SSL/1.48
Cache-Control: no-store, no-cache, must-revalidate, post-check=0,
pre-check=0
Pragma: no-cache
Set-Cookie: PHPSESSID=2989258979bd07405999448563ef4bfc; path=/
Expires: Thu, 19 Nov 1981 08:52:00 GMT
Content-Length: 2809
Content-Type: text/html
------------------------------
The real content length is 2890 bytes.
I make use of the following PEAR libraries:
HTML_Quickform, ADODB session management, and Log (when enabled). I
make use of mySQL and ctype() functions throughout my application. I
utilise a "user-supplied session function" for PHP to make use of
ADODB as opposed to using the built-in default session handler.
=Apollyon=
------------------------------------------------------------------------
[2003-08-20 20:34:39] phpbug at paypc dot com
Testing with that script fragment (as the entire web-page), I get the
following for output:
------------ SNIP -------------
HTTP/1.1 200 OK^M
Date: Thu, 21 Aug 2003 01:28:13 GMT^M
Server: Apache/1.3.27 Ben-SSL/1.48^M
Connection-Type: Keep-Alive^M
Content-Length: 0^M
Content-Type: text/html^M
^M
------------ SNIP -------------
I assume this is correct. I displayed the CR's (I'm on UNIX).
Apache 1.3.x doesn't support HTTP/1.1 style connections?
In Apache 1.1 and 1.2 documentation, I see:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Enabling Keep-Alive Support
Apache 1.1 comes with Keep-Alive support on by default, however there
are some directives you can use to modify Apache's behavior:
Note: Apache 1.2 uses a different syntax for the KeepAlive directive.
<<<<<<<<<<<<<<<<<<<<<<<<<<<<
I am *NOT* using PHP built as a CGI SAPI - it's a statically linked
MODULE built into Apache-SSL.
=Apollyon=
------------------------------------------------------------------------
[2003-08-20 10:32:07] [EMAIL PROTECTED]
What if eliminating the following line?
header("Connection-Type: Keep-Alive");
Note that HTTP/1.1 style connection is not supported in either Apache 1
or CGI SAPI.
And also try to generate a null output and see how it'll work:
<?php
header("Connection-Type: Keep-Alive");
header('Content-Length: 0');
?>
------------------------------------------------------------------------
[2003-08-20 07:05:31] phpbug at paypc dot com
Description:
------------
Very straightforward.
I have a preamble and postamble include on ALL of my main .php pages...
which use ADODB session management, some PEAR, and other typical
dynaweb goodies.
The Content-Length header is *ALWAYS* off by about 81 bytes, as in, the
ob_get_length() is always smaller than the actual size of the content
sent...
I've verified this with "netcat" - after trimming the headers from the
reply, the content is 81 bytes larger than advertised in the
Content-Length header.
I observed no change in behaviour after adding the ob_clean() call into
my header file.
OTHERWISE, things work fine.
What's the deal? Is Apache silently adding CRLF's?
I am NOT using the gzhandler or transparent compression here. This is
just straightforward output.
=Lord Apollyon=
Reproduce code:
---------------
The code for all of my index.php's consist of the following:
<?PHP
include_once 'db_header.inc';
................
include_once 'db_footer.inc';
?>
db_header.inc:
<?php
$start_time=gettimeofday();
ob_clean(); // paranoid and desperate
ob_start();
header("Connection-Type: Keep-Alive");
?>
db_footer.inc contains:
<?PHP
if ($LOGGING)
{
$LOG->close();
$LOGGING=FALSE;
}
header('Content-Length: ' . ob_get_length());
ob_end_flush();
?>
Note: there are no blank lines at the beginning of my *.inc or *.php
files.
Expected result:
----------------
I would expect Content-Length to reflect the actual output size.
------------------------------------------------------------------------
--
Edit this bug report at http://bugs.php.net/?id=25175&edit=1
