Question

I want to log access to any files in the /files folder, so I can process it with PHP to generate some statistics.

I don’t want to write a custom PHP handler called via RewriteRule because I don’t want to have to deal with status codes, MIME-types and caching headers, and file locking issues.

I don’t have access to the server configuration, so I can’t use CustomLog (I do have access to .htacess).

I can’t use X-Sendfile because it’s not enabled.

I don’t have access to the access.log.

---

Looking for an authorative answer.

Answer 1

I’ve tried a great many things and there seems to be no easy solution.

My solution uses the Location header trick proposed by @yes123 but I’ve tweaked it to match my preferences.

The links to the files are kept intact, so it’s still: /files/path/to/my/file.abc
I have a RewriteRule:

RewriteRule ^files/(.*) path/to/tracker.php?path=/$1

Then in the file I issue a Location header by adding ?track=no to the URL and an exception to the earlier RewriteRule:

RewriteCond %{QUERY_STRING} !(&|^)track=no(&|$)

I’ve added one more optimization. I’ve enabled E-Tags so if the client send an E-Tag header, see if it matches the file and return a 304 Not Modified instead of a Location.

$fs = stat($document_root . $path);
$apache_etag = calculate_apache_etag($fs);
if ((isset($_SERVER["HTTP_IF_MATCH"]) && etag_within_range($_SERVER["HTTP_IF_MATCH"], $apache_etag))
    || (isset($_SERVER["HTTP_IF_NONE_MATCH"]) && etag_within_range($_SERVER["HTTP_IF_NONE_MATCH"], $apache_etag))
) {
    header("ETag: " . $apache_etag, true, 304);
    exit;
}

function etag_within_range($etag1, $etag2) {
    list($size1, $mtime1) = explode("-", $etag1);
    list($size2, $mtime2) = explode("-", $etag2);
    $mtime1 = floor(hexdec($mtime1) / 1000000);
    $mtime2 = floor(hexdec($mtime2) / 1000000);
    return $mtime1 === $mtime2 && $size1 === $size2;
}

And implementation for calculate_apache_etag can be found here: How do you make an etag that matches Apache?

etag_withing_range solves the issue of comparing against a higher precision mtime in Apache.

---

Notes on solutions that didn’t work

virtual

Test script:

var_dump(apache_response_headers());
virtual("/path/to/image.jpg");
var_dump(apache_response_headers());

Outputs:

array(1) { ["X-Powered-By"]=> string(10) "PHP/5.2.11" }
[[binary junk]]
array(5) { ["X-Powered-By"]=> string(10) "PHP/5.2.11" ["Keep-Alive"]=> string(18) "timeout=5, max=100" ["Connection"]=> string(10) "Keep-Alive" ["Transfer-Encoding"]=> string(7) "chunked" ["Content-Type"]=> string(9) "text/html" }

Content-Type: text/html reaaaaalllly? 🙁

Perhaps PHP5.3’s header_remove function can solve this? I haven’t tried.