Designing a test for gzip performance

Hi,

The Mobile Web Best Practices Working Group discussed last week [1]
whether or not to advertise the use of GZIP-encoding as being a good
practice to serve content to mobile devices.

The crux of the question is to find the right balance between:
 * the time saved by having less data transmitted on mobile networks
(with sometimes a limited bandwidth)
 * the time taken to decompress the resulting resource on mobile phones
(with a limited CPU)

I've started to write a test that tries to assess these two aspects:
http://www.w3.org/2008/06/gzip-mobile/

What the test does:
 * as soon as the server receives the request for the first step of the
test, it records a log point in a database
 * when the client finishes to parse the 56K page (full of comments,
mostly), it hits a remote javascript (called from <script> at the end of
the document)
 * when the server receives the request for the script, it logs a new
point in the database
 * then the same sequence of operations is made for roughly the same
content, but sent g-zipped

It gathers results as follows:
http://www.w3.org/2008/06/gzip-mobile/results

I'm interested to hear whether you guys think this is a valid approach
or not, how to possibly improve it or make it more complete.

Thanks for any feedback,

Dom

1. http://www.w3.org/2008/06/17-bpwg-minutes#item_conservative