One of my personal coding standards is to keep similar HTML attributes grouped together in a consistent order at the start of the element, e.g.

1
2

<script type="text/javascript" src="foo"></script>
<script type="text/javascript" src="bar"></script>

vs.

1
2

<script src="foo" type="text/javascript"></script>
<script src="bar" type="text/javascript"></script>

This helps with readability, maintainability, &c. &c.

Recently another thought occurred to me: that this would also help with HTTP compression.

I ran a quick test using these two (very simplified) HTML documents:

With the following results (taken from the Firebug Net panel):

So, the ordered version gives us a saving of 3 bytes – not a massive gain, but in a much larger document it could be significant.

I imagine this would also hold true (and possibly give much more significant savings) for CSS if property/value pairs that are the same are grouped together.

For those of you interested in such things, I tested the compressed vs. uncompressed HTML by first soft linking the HTML files:

ln -s ordered.html ordered-uncompressed.html
ln -s unordered.html unordered-uncompressed.html

The adding the following .htaccess:

AddOutputFilterByType DEFLATE text/html
SetEnvIfNoCase Request_URI uncompressed no-gzip dont-vary

which disables output compression for any URI that contains ‘uncompressed’.