[buro9]

Originally Posted by AlexanderT

David, can you show me examples of your findings? I generally don't believe that the compression ratio is that much affected by the mere presence of whitespace. But of course I could be wrong

Hehe, that was a nice challenge, but one I enjoyed.

To offer a proof I've done the following:
wget my forum index.php both compressed using the now default non-aggressive regexp, and also without this hack applied.

The sizes of the two files:
37,732 bytes = Compressed file
41,348 bytes = Uncompressed file

A 9% reduction in filesize.

I then ran gzip from within SSH, this should be the same as PHP gzip'ing them or a close approximation thereof:

6,748 bytes = Compressed + GZIP
8,102 bytes = Uncompressed + GZIP

A 20% reduction in final filesize.

The additional reduction of the size is higher than merely the white space being stripped and is an indicator that doing this does offer a benefit to GZIP.

So the compressed page (exactly the same HTML but without white space) ends up being 20% lighter with GZIP'ing than the non-compressed page.

Of course, the mileage varies on every page and with each style that you use. I should point out that the above was run against my current forum homepage which is not the default vBulletin style and that I've hacked my page quite a lot. So the filesize of your page and the amount it compresses and the amount it GZIPs are all factors depending on your configuration and the amount of whitespace in a page, etc.

The worst I've seen from this hack is a mere 3% reduction in file size on some of the smaller pages (such as the error messages).

The biggest benefit isn't the saved 1k of traffic... but the speed at which the rendering increases. Whether that is because the file transfered quicker, of the DHTML engine didn't have to work so hard ignoring whitespace, etc... I simply do not know... but it does appear to render faster and that is my prime goal