ob_gzhandler

(PHP 4 >= 4.0.4, PHP 5, PHP 7)

ob_gzhandler — 在ob_start中使用的用来压缩输出缓冲区中内容的回调函数。ob_start callback function to gzip output buffer

说明

ob_gzhandler ( string $buffer , int $mode ) : string

ob_gzhandler()目的是用在ob_start()中作回调函数，以方便将gz 编码的数据发送到支持压缩页面的浏览器。在ob_gzhandler()真正发送压缩过的数据之前，该 函数会确定（判定）浏览器可以接受哪种类型内容编码（"gzip","deflate",或者根本什么都不支持），然后 返回相应的输出。 所有可以发送正确头信息表明他自己可以接受压缩的网页的浏览器，都可以支持。 All browsers are supported since it's up to the browser to send the correct header saying that it accepts compressed web pages. 如果一个浏览器不支持压缩过的页面，此函数返回FALSE。

参数

buffer

mode

返回值

更新日志

版本 说明
4.0.5 填加了 mode 参数。

范例

Example #1 ob_gzhandler() example

<?php

ob_start("ob_gzhandler");

?>
<html>
<body>
<p>This should be a compressed page.</p>
</html>
<body>

注释

Note:

ob_gzhandler() 需要 zlib 扩展。

Note:

不能同时使用ob_gzhandler() 和 zlib.output_compression。 也要注意使用 zlib.output_compression 要优于 ob_gzhandler()。

参见

User Contributed Notes

Jer 16-Apr-2010 04:17
I've just spent 5 hours fighting a bug in my application and outcome is:

<?php
// do not use
ob_start("ob_gzhandler");
// along with
header('HTTP/1.1 304 Not Modified');

// or in the end use
ob_end_clean();
?>

W3C Standart requires response body to be empty if 304 header set. With compression on it will at least contain a gzip stream header even if your output is completely empty!

This affects firefox (current ver.3.6.3) in a very subtle way: one of the requests after the one that gets 304 with not empty body gets it response prepended with contents of that body. In my case it was a css file and styles was not rendered at all, which made problem show up.
przemekryciuk at gmail dot com 23-Jun-2009 02:34
The simplest way for gzip compression is:
<?php
if(!ob_start("ob_gzhandler")) ob_start();
?>
ob_start("ob_gzhandler") returns FALSE if browser doesn't support gzip, so then is called normal ob_start();
walter 11-Dec-2008 07:48
I have a very well tested module in my framework (ie: running in production environments for a couple of years) called 'fastpage', which takes care of both memcaching and gzipping for frequently requested content.  It memcaches one version of each unique page both gzipped and non-gzipped, and returns the appropriate version based upon the browser's Accept-encoding header.

Use is:
$page_spec = array($dependent,$variables,$go,$here);
if(!fastpage_display('content_id',$page_spec)) {
 ... make page here ...
 fastpage_displayed('content_id',$page_spec);
}

Unfortunately, on IE8 beta (in IE8 mode, or in IE7 emulation mode), the fastpage-enabled content was not being un-gzipped.  IE would request the content, a 200 status and correct byte count would appear in the server logfile, and IE would show no errors.  However, CSS would not be applied and Javascript would not execute.

The only fix I can find right now is to disable gzip entirely for this browser.  User agent string is: Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 5.1; Trident/4.0; .NET CLR 2.0.50727

The same output works fine in firefox (windows, linux, mac), and in safari.

Versions are: zlib 1.2.3-r1 / php 5.2.6-r7 / lighttpd 1.4.20

So much for the 'this has been fixed in IE7' comment, above...
ronald at hidden dot com 26-Aug-2008 01:15
Hi,

if you are using apache, I think you are much better of using apache's compression, as this will also compress static text files (css, js, html, etc) and it will keep your php-code clean.

See http://httpd.apache.org/docs/2.0/mod/mod_deflate.html for an excellent how to.

And for other webservers, I'm pretty sure they have equivalents.

Cheers, Ronald
sugoi82 at removeme dot gmail dot spam dot com 24-Dec-2007 06:09
in reply to nyarl at hotep dot co dot uk;

Your 'solution' doesn't really work since session_id() will return an empty string if there is no session started. Therefore it will never return NULL.
Yes the problem is gone, but also is your output buffering xD

I haven't found a solution to this problem yet... It seems ie6 just doesn't wanna play nice - like it does most of the times.
nyarl at hotep dot co dot uk 09-Dec-2007 06:45
Seems others have had problems with IE6 giving blank pages erratically when using ob_start("ob_gzhandler"). One reason may be that a session has already been started, so try:

<?php
  if (session_id() === null ) ob_start('ob_gzhandler');
?>

Worked for me. :)
TheSeoDude 09-Sep-2007 03:34
I wrote a simple script that I use on my site. Bandwidth consumption dropped dramatically. Pages shrunk 3-5 times.

The script simulates the results of using this function directly but by using regular output buffering with ob_start. So you get to output gzip and play with it before you do.

You can find the script with explanations here:
http://www.tellinya.com/art2/106/gzip-compress-html

Cheers and input is always welcome fellow coders!
maratd at gmail dot com 29-Aug-2007 08:51
Someone previously mentioned that MSIE doesn't cache gzipped content. This is false. He misread the source of information. In fact, IE will cache gzipped content no matter what. Here is the quote from the mailing list:

The reason the COMPRESSED responses are, in fact,
always getting cached no matter what "Vary:" field name
is present is just as I suspected... it is because MSIE
decides it MUST cache responses that arrive with
"Content-Encoding: gzip" because it MUST have a
disk ( cache ) file to work with in order to do the
decompression.

Source: http://lists.over.net/pipermail/mod_gzip/2002-December/006826.html
tehjosh at gamingg dot net 01-Aug-2007 04:18
faisun at sina dot com:
If you read up on output buffering, you'll see that if an output buffer callback returns false, this instructs PHP to output the original string untouched.  That's why ob_gzhandler() returns false if the encoding is not supported.   When ob_start("ob_gzhandler") is used and encoding is not supported, ob_gzhandler() will return false and PHP outputs the original, uncompressed string.

jsnell at e-normous dot com:
Output buffering callback functions accept up to two parameters, so this would probably work better for your situation:

<?php
function ob_gz_handler_no_errors($buffer, $mode)
{
    @ob_gzhandler($buffer, $mode);
}

ob_start('ob_gzhandler_no_errors');
?>

However, if you're trying to suppress errors caused by headers already being sent, it would be better to start the output buffering earlier, before any output can be sent.
jsnell at e-normous dot com 31-May-2007 11:04
If you want to suppress errors (such as for cases when headers are already sent) and you don't want to write or patch your error_handler, you need to create an intermediate callback:

function ob_gz_handler_no_errors($buffer)
{
    @ob_gzhandler($buffer);
}

ob_start('ob_gzhandler_no_errors');
sredhead 27-Feb-2007 03:35
One of the earlier posts says caching of gzipped files is broken in all versions of MSIE. The following blog entry appears to declare that this has been fixed in IE7, but I haven't verified this.

http://blogs.msdn.com/ie/archive/2005/10/31/487509.aspx
nicolas dot grekas+php at gmail dot com 28-Jan-2007 10:45
Here are some precisions :

- the "mode" arg accepts a bit field composed of PHP_OUTPUT_HANDLER_START, PHP_OUTPUT_HANDLER_CONT and PHP_OUTPUT_HANDLER_END. See http://www.php.net/manual/fr/ref.zlib.php#56216 for an example. The value that jazfresh recommends below (5) is the good one, because 5 == PHP_OUTPUT_HANDLER_START | PHP_OUTPUT_HANDLER_END.

- when called INSIDE an output buffering handler, ob_gzhandler DOES NOT return false when the browser doesn't support compressed pages. It rather returns the original string.
fpiat at bigfoot dot com 27-Aug-2006 05:55
DON'T compress/gzip your Error pages :

i can think of two good reasons not to compress your error pages:
1. You want to make sure that people can read you error pages, don't you ?
2. Compressing pages have a "side effect" : it's smaller !! if you return "HTTP/1.x 404 Document Not fount" (or any other HTTP error), MSIE will display it's built in error message, not yours if it's small than 513bytes. (see "HOW TO: Turn Off the Internet Explorer 5.x and 6.x 'Show Friendly HTTP Error Messages' Feature on the Server Side" http://support.microsoft.com/default.aspx?scid=kb;en-us;Q294807 )
spam_this at zleelz dot com 26-Aug-2006 11:49
When writing scripts for distribution, I would usually "null" out the following deprecated superglobals so that users who uses the script will not be able to use them.
<?php
$HTTP_GET_VARS = null;
$HTTP_POST_VARS = null;
$HTTP_COOKIE_VARS = null;
$HTTP_POST_FILES = null;
$HTTP_SERVER_VARS = null;
?>

However, when using ob_start('ob_gzhandler'), one of those superglobals somehow disable this function.

Found out that it was $HTTP_SERVER_VARS that's causing the problems.

<?php
ob_start('ob_gzhandler');

$HTTP_GET_VARS = null;
$HTTP_POST_VARS = null;
$HTTP_COOKIE_VARS = null;
$HTTP_POST_FILES = null;
/**
 * Causing the trouble
 */
//$HTTP_SERVER_VARS = null;
?>

I don't know why it does that, but I just want to point that out.
richard at mf2fm dot com 11-Jul-2006 02:50
According to the manual, if ob_gzhander detects that the browser is unable to support deflate or gzip it returns FALSE.  Does this mean that if you use ob_start("ob_gzhandler"), any browsers that do not support gzip/deflate will receive a blank page?

I've been having problems with IE6 displaying gzipped pages and adding a test:

<?php

if (substr_count($_SERVER['HTTP_ACCEPT_ENCODING'], 'gzip')) ob_start("ob_gzhandler");

else ob_start();

?>

solved the problem.  Surely if ob_gzhandler detects a lack of support for deflation it should return the input unchanged and not FALSE?
michael at michaelphipps dot com 29-Apr-2006 03:34
It is also possible to use ob_gzhandler to compress css and javascript files, however some browsers such as firefox expect content type text/css on css files.

To get around this send a content type header:

<?php
ob_start('ob_gzhandler');
?>

.... your css content ...

<?php
header("Content-Type: text/css");
header("Content-Length: ".ob_get_length());
ob_end_flush();
?>
mariusads[at]helpedia[dot]com 04-Feb-2006 09:56
Please be aware that, as other users have already mentioned, the compression will fail if there are characters before the php start tag "< ?".

This is also the case when saving files in UTF-8 format with editors such as Ultraedit. Make sure you save the file using the defined option UTF-8 NO BOM or delete the BOM otherwise the two UTF BOM characters will be written at the start of the file.
lapoint2 at msu dot edu 06-Sep-2005 06:40
So why doesn't the web server just do this by default? 
This works for me if I do:
ini_set('zlib.output_compression_level', 3);  ob_start("ob_gzhandler");
or even just:
ob_start("ob_gzhandler");

I did the level 3 compression, I think the default was 6 and I didn't want to put too much load on the server.  For a 895k file (my largest) the compression levels were:
1 = 189k
3 = 178k
4 = 163k
6 = 156k    (I believe 6 is the default if you leave out the ini_set)
9 = 155k
I use http://www.whatsmyip.org/mod_gzip_test/ to check the sizes.

FYI: This works with dynamic files in Movable Type 3.x (I'm testing it in 3.2) I've got the above command in the first line of my mtview.php file.

More info here: http://www.whatsmyip.org/forum/viewtopic.php?t=43

On several sites I've read the some browsers don't like compressed CSS.
Tony Piper 14-Jun-2005 07:11
I just spent 30 minutes wondering why my browser wasn't getting the gzipped version :-O

If you're using Google Web Access your pages will  be delivered to the browser uncompressed unless you tell GWA to ignore the specific site.

Obvious, really.
24-May-2005 03:27
No matter the circumstances, you MUST NOT check the user agent string to find out if gzip compression is enabled. HTTP defines the Accept-Encoding header for exactly this purpose and you MUST check it before enabling compression.
m at rtij dot nl 16-Apr-2005 05:29
All versions of MSIE have a bug where they don't cache gzipd contents. If your application depends on caching, this is something to keep in mind. In the end, I did:

<?php

// These are so benificial, they default to true.
if (!isset($use_page_cache))
    $use_page_cache = 1;
if (!isset($use_compression))
    $use_compression = 1;

// Add browsers here as we must detect them. Opera is an oddball, if we don't detect
// it specifically, it will turn up as MSIE

$browser="other";
if (isset($_SERVER['HTTP_USER_AGENT'])) {
    $agent = $_SERVER['HTTP_USER_AGENT'];
    if (eregi("opera",$agent)){
    $browser="opera";
    }elseif(eregi("msie",$agent)){
    $browser="msie";
    }
}

if ($use_compression && !( $use_page_cache && $browser == "msie")) {
    // Turn on compression, makes quite a difference in bandwith usage.
    // However, MSIE (all versions) have a bug with compression and caching. So for MSIE
    // it's either compression or caching. We choose caching.
    ob_start('ob_gzhandler');
}

session_cache_limiter("must-revalidate");

session_start();

// ... put stuff in $content ...

if ($use_page_cache) {
    // MD5 is slow, however with a fast server (PIII or better) we should be OK
    $hash = md5($content);
    $headers = getallheaders();
    if (isset($headers['If-None-Match']) && ereg($hash, $headers['If-None-Match']))
    {
    header('HTTP/1.1 304 Not Modified');
    exit;
    }
    header("ETag: \"$hash\"");
}

print $content;

?>
grange (at) club-internet (dot) fr 07-Mar-2005 07:17
It may seem obvious but if you use ob_start("ob_gzhandler"),  make sure no content is unintentionally echo'ed before : this could be blank lines before '<?php' or after '?>' tags in included files or even errors.

Otherwise, only a part of the content (the content sent after ob_start()) will be compressed, which will confuse the client.

Setting compression using zlib.output_compression either in php.ini or in Apache configuration file (with directive php_flag) is safer in this regard.
wallacebw at some-domain dot com 21-Feb-2005 09:50
IE 6+ appears to validate the CRC and disgard invalid gzip content FYI...
Brian
andy dot rem dot ove dot this at occ dot be 01-Apr-2004 02:21
From what I can understand, quite some people seem to have problems with this functionality.  If you want to do this quick, like i did, you can use the outstanding class gzip_encode made by Leknor at http://leknor.com/code/php/view/class.gzip_encode.php.txt
mazsolt at yahoo dot com 18-Mar-2004 07:20
if you want to send an output to the browser (which accepts gzip), and you haven't set the buffering callback ob_start("ob_gzhandler"), you may use the gzencode() function.

header("Content-Encoding: gzip");
echo gzencode("some output", 9, FORCE_GZIP);
daijoubuNOSP at Mvideotron dot ca 27-Oct-2003 02:05
So much code...here's a lighter/faster one:
<?php

if ( $do_gzip_compress )
{
    $gzip_size        = ob_get_length();
    $gzip_contents    = ob_get_clean(); // PHP < 4.3 use ob_get_contents() + ob_end_clean()
   
    echo "\x1f\x8b\x08\x00\x00\x00\x00\x00",
         substr(gzcompress($gzip_contents, 9), 0, - 4), // substr -4 isn't needed
         pack('V', crc32($gzip_contents)),    // crc32 and
         pack('V', $gzip_size);               // size are ignored by all the browsers i have tested
}

exit;
?>
I wouldn't recommande using level 9, waste of CPU, something between 1 and 6 would be more resonable
I have tested without the substr/crc/size and it worked flawlessly in IE5/5.5/6, Opera7.11/7.21, IE5.5 MAC
brewthatistrue at plzYnoAspamHorOevilO dot com 21-Sep-2003 02:04
i had a LOT of trouble trying to get gzipping working correctly
i kept getting a blank page on the first load, although subsequent pages loaded fine.
i tried stuff with ob_flush and ob_end_clean and all kinds of stuff i didn't understand. i'm totally confused about different versions of php and different methods of gzipping.

in the end, i got it working (maybe i'll understand it later).
i ended up using phpBB's code (stripped of non-gz stuff, and ever-so-slightly modified)
http://phpbb.com/

index.php:
------------
<?php
require_once('top.php');

echo "<html>\n<head>\n<title>Gzip Test</title>\n<body>\n<h1>testing</h1>\n</body>\n</html>";

require_once('bottom.php');
?>

gz_header.php - taken form phpBB's page_header.php
--------------
<?php
    $phpver = phpversion();

    $useragent = (isset($_SERVER["HTTP_USER_AGENT"]) ) ? $_SERVER["HTTP_USER_AGENT"] : $HTTP_USER_AGENT;

    if ( $phpver >= '4.0.4pl1' && ( strstr($useragent,'compatible') || strstr($useragent,'Gecko') ) )
    {
        if ( extension_loaded('zlib') )
        {
            ob_start('ob_gzhandler');
        }
    }
    else if ( $phpver > '4.0' )
    {
        if ( strstr($HTTP_SERVER_VARS['HTTP_ACCEPT_ENCODING'], 'gzip') )
        {
            if ( extension_loaded('zlib') )
            {
                $do_gzip_compress = TRUE;
                ob_start();
                ob_implicit_flush(0);

                header('Content-Encoding: gzip');
            }
        }
    }
?>
gz_footer.php - taken form phpBB's page_tail.php
------------
<?php
// Compress buffered output if required and send to browser
if ( $do_gzip_compress )
{
    //
    // Borrowed from php.net!
    //
    $gzip_contents = ob_get_contents();
    ob_end_clean();

    $gzip_size = strlen($gzip_contents);
    $gzip_crc = crc32($gzip_contents);

    $gzip_contents = gzcompress($gzip_contents, 9);
    $gzip_contents = substr($gzip_contents, 0, strlen($gzip_contents) - 4);

    echo "\x1f\x8b\x08\x00\x00\x00\x00\x00";
    echo $gzip_contents;
    echo pack('V', $gzip_crc);
    echo pack('V', $gzip_size);
}

exit;
?>

i must admit to not seeing this working in opera 7.11. perhaps i will figure that one out.
if there is anything i should change email me and i can edit my post
jazfresh at SPAM-JAVELIN dot hotmail dot com 18-Sep-2003 07:54
In the set_error_handler notes, there is a technique described for capturing all errors (even parse errors) before they are displayed the user, using a special error handler and output handler. If this output handler detects a fatal error in the output buffer, it's captured and dealt with before it can be displayed to the user. If no error was detected, then output buffer is displayed verbatim (i.e. without being compressed).

If you are using this method, you can still take advantage of ob_gzhandler's compression. However, you MUST specify a mode argument (I'm using 4.2.2 on RedHat9). The mode value affects which headers are automatically added (Content-Encoding, etc). A value of '5' worked for me. '0' or discarding the argument produces a blank screen under Mozilla.

<?php

function my_output_handler(&$buffer) {
  // Detect errors in the output
  if(ereg("(error</b>:)(.+) in <b>(.+)</b> on line <b>(.+)</b>", $buffer, $regs)) {
    my_error_handler(E_ERROR, $regs[2], $regs[3], $regs[4]);
    // ...
    // ... Insert your error handling here ...
    // ...
    return 'An internal error occurred.';
  } else {
    // The page rendered without any errors, so compress
    // and output.
    return ob_gzhandler($buffer, 5);
  }
}
?>
daijoubu at videotron dot ca 06-Sep-2003 04:35
About the previous note from Davey:

ob_start(array('ob_gzhandler',9));

Does not work. The output size isn?t affected at all, stays the same.

ob_gzhandler compression level use zlib.output_compression_level, which is -1 per default, level 6.

To change the compression level on the fly, simply use ini_set:
<?php
ini_set('zlib.output_compression_level', 1);
ob_start('ob_gzhandler');
echo 'This is compressed at level 1';
?>
09-Apr-2003 10:47
When using ob_start("ob_gzhandler"), be aware that the output buffer has to be flushed in order to invoke the ob_gzhandler callback function.

This may not always happen. For example, if you use ob_get_contents() to copy the output buffer to a string for further manipulation and then silently discard the buffer with ob_end_clean(), the output buffer is never "flushed" and as a result, the ob_gzhandler callback function is never invoked. And your pages won't be compressed.

This will be the case, for example, if you use the PHP Fusebox architecture/framework.
davey at its-explosive dot net 03-Apr-2003 06:41
To pass the second argument to ob_gzhandler() which specifies what level of compression should be used (which I assume is 1-9 as with the gzip binary, with 9 using the most processor and time to do, 4 is the standard IIRC) you must call ob_start() like so:

ob_start(array('ob_gzhandler',9));

- Davey
nospam at 1111-internet dot com 08-Mar-2003 03:23
Differences between zlib.output_compression and ob_gzhandler:

zlib.output_compression runs in parallel with script execution - it begins compressing as soon as it receives any output from your script, and sends data to the client each time its buffer (4K by default) gets full. ob_gzhandler (actually 'ob_start("ob_gzhandler");') will not start compressing until the script flushes (or, usually, exits), and will in turn send the entire compressed document at once - which makes it more susceptible to causing a perception of latency.

Advantage: zlib.output_compression

On the other hand, ob_gzhandler gives you script-level control allowing you to use it selectively or to unset it after setting it in certain cases. Despite some documentation to the contrary, zlib.output_compression does not appear to be able to be set or unset on a script level; you must instead set it globally (in php.ini) or in your webserver configs or .htaccess files, possibly using FilesMatch-type mechanisms to control which scripts it will or won't apply to - which could get unwieldy for large projects - particularly those which employ PHP to produce images and other non-text output in addition to normal text output.

Advantage: ob_gzhandler

Net advantage: depends on your particular needs. I'm trying zlib.output_compression for now, but I miss the flexibility that ob_gzhandler would provide.

p.s. here's an Apache 1.3.* httpd.conf/.htaccess file snippet that demonstrates the syntax for conditionally setting zlib.output_compression:

<FilesMatch "\.(php|html?)$">
# turn it on with the buffer set to 2K using php_value
php_value zlib.output_compression 2048
# or just turn it on using php_flag
# php_flag zlib.output_compression On
</FilesMatch>
aki at robotnik dot net 04-Mar-2003 11:13
if you get errors like:
"output handler 'ob_gzhandler' cannot be used twice"
if you are using
"ob_start("ob_gzhandler");"

check for your php.ini file, it should look like this:

output_buffering = Off  ; delete the 4096k value
output_handler =
zlib.output_compression = Off
dsugar100 at dolphinsoft dot com 06-Nov-2002 03:33
I have noticed that if you have the php setting for zlib.output_compression set to 'On' AND you attempt to use hte ob_gzhandler handler to ob_start () the output you get in the browser WILL be garbled in PHP 4.2.3.  I'm guessing that the output buffering is compressing the output to be sent, and then zlib is doing it a second time, but the browser is only uncompressing it once.

But using either one will give you the same results (compressed output from the script)
nid at home dot se 30-Sep-2002 04:10
If you are trying to use the ob_gzhandler in a MSHTA (Microsoft HTML Application) you will find that your pages does not show up. the MSHTA-browser probably says it can handle gzip, but it just won't work (IE:6, PHP:4.1.2)
psychones at ifrance dot com 08-Aug-2002 05:18
Problem revelate by xn@bnw.com

cause real problem when you have a global call
to ob_gzhandler , for example in an include configuration
file , and you want send a non gzip content , for any reason ...

To resolve it ,

put this after your global call to ob_gzhandler

include("conf_that_call_ob_gzhandler.php");
-->ob_end_clean();
-->header("Content-Encoding: none");

If you try to overload the header before the the ob_gzhandler call
or during the buffering output , it will not work .
( probably erase at the call of ob_gzhandler  and not allow during
buffering ).

Hope this help
isoma at altavista dot net 24-Jun-2002 07:44
RFC 2616 suggests that a more correct way to compress documents on the fly is to use gzip transfer encoding rather than gzip content encoding.    However, it's perhaps not advisable to use this since client support for it is extremely limited.

The difference is most evident when you come to save a file from your browser.  If it's gzip content encoded then the browser should (and may well) save it gzipped.  If it's gzip transfer encoded then the browser should uncompress it first.
Sven dot Dickert at planb dot de 29-May-2002 04:54
Following Browsers send
Accept-Encoding: gzip
but have Problems with
Content-Encoding: gzip

* netscape < 5 have problems with CSS/Javascript parts sendet in gzip
* opera 5.x,6 store downloads compressed
* lynx 2.8.4 store downloads compressed but tells this with attaching ".gz" to the filename

You should know about:
* Internet Explorer 5.x,6 should turned to HTTP 1.1 to get "Accept-Encoding: gzip"
* In Mozilla 1.0 RC2 you can write the "Accept-Encoding:" in Preferences/Advanced/HTTP Networking
xn at bnw dot com 12-Apr-2002 08:40
if you call ob_end_clean() after ob_start("ob_gzhandler"), the "Content-Encoding: gzip" header will still get sent (assuming the browser supports the encoding).  if you don't call ob_start() again with the ob_gzhandler callback function, the output will not be compressed, but the header will say it is.  this causes mozilla (as of build 2002032808) to display a blank page.
junior at jaj dot com 04-Mar-2002 10:12
To get this to work properly, you have to compile PHP with "--with-zlib". If you don't, you won't get any errors, it just won't actually compress anything. This is a phenomenal thing. For just a small bit of processor time, you can DRASTICALLY reduce the bandwidth requirement of your scripts. There are very few circumstances in which this should not be used.