I have a script that generates my XML sitemap and writes it to the file sitemap.xml.gz - i.e. an XML file, compressed with gzip. This file is definitely written correctly as when I download it via FTP it's all good.

However, when I download the file direct from the site (over HTTP), the resulting file appears to be doubly-compressed. When I unzip the file, the sitemap.xml file is a binary file. If I rename that to sitemap2.xml.gz and try unzipping again, I get the true XML file.

So I think the server (Apache2) is for some reason taking the .gz file and serving it with gzip compression again. The headers for the file come back as this:

Status: HTTP/1.1 200 OK
Date: Mon, 16 Jul 2012 00:00:47 GMT 
Server: Apache  
Last-Modified: Sun, 15 Jul 2012 23:35:26 GMT    
ETag: "89fff2-3bc46-4c4e6c48deb80"  
Accept-Ranges: bytes    
Vary: Accept-Encoding   
Content-Encoding: gzip  
Connection: close   
Transfer-Encoding: chunked  
Content-Type: application/x-gzip

In my httpd.conf I have this:

# compress all text & html:
AddOutputFilterByType DEFLATE text/html text/plain text/xml application/xml text/css text/javascript application/x-javascript application/javascript

My VirtualHost declaration only has some mod rewrite stuff.

Anyone have any ideas why Apache might be sending the gzip header for this file?

UPDATE: I removed the application/xml entry from the AddOutputFilterByType line, and the file now downloads normally like any other binary file. However, the problem now is that regular .xml files are no longer sent gzipped.

So it seems like the server is deciding that .xml.gz files should be parsed as application/xml, even though it sends it with the header application/x-gzip.

Also, I checked the /etc/mime-types file, it doesn't have an entry for gzip and has this comment at the top:

Note: Compression schemes like "gzip", "bzip", and "compress" are not actually "mime-types". They are "encodings" and hence must not have entries in this file to map their extensions. The "mime-type" of an encoded file refers to the type of data that has been encoded, not the type of encoding.

  • For sanity/just to see: What happens if you take a plain file and just name it sitemap.xml.gz?
    – Su'
    Commented Jul 16, 2012 at 0:44
  • It's very hard to be certain about the answer here. Do you have a .htaccess file that's altering the behavior ? Is the behavior consistent in different browsers ? (wget, ff, chrome, ie) I would have thought that downloading the file would at the worst double compress it with the browser uncompressing and saving the sitemap.xml.gz single compressed.
    – xyious
    Commented Jul 16, 2012 at 1:19
  • @xyious no htaccess files but the rules that would normally be in htaccess are in sites-enabled/mysite.conf in a VirtualHost declaration. As mentioned above, they are only rewrites/redirects, nothing that would affect encoding. I tested in several browsers with the same result. Commented Jul 16, 2012 at 13:40
  • @Su' if I just rename a plain text file (3KB) to file.txt.gz and download it via the browser, it appears to end up as a regular gzip file (1.4KB) containing the original text file (3KB). Commented Jul 16, 2012 at 13:48
  • 1
    Based on your edit, I ran across a ServerFault question that seems to match(with solution). See if that works out for you.
    – Su'
    Commented Jul 16, 2012 at 13:59

1 Answer 1


Quoting @cyberx86 over at ServerFault (who you should go and vote up):

The .xml.gz filetype may be defined as being an xml file (e.g. with forcetype in a filesmatch block) - which would cause Apache to match it to the type above.

I think you can get around that by adding an exception, above it:

SetEnvIfNoCase Request_URI ".xml.gz$" no-gzip dont-vary

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.