(1) would break all existing stylesheets. Although the present format may not be ideal (?), it is working.

Breaking code just for beauty's sake is a very bad thing!

According to discussion on the mailing list, custom stylesheet are seldom therefore not so many users would be impacted.

Additional motivation for these changes: current xsl takes huge time (hours) on very large result files (~100 webtest à 1000 steps) to generate the html reports. Furthermore recent improvements of the standard xsl have made this task even more time intensive.

I lost a lot of time (and that means weeks, not hours) changing existing tests to follow the changes we foolhardily made to webtest.

Staying compatible with previous versions should belong to our goals.

The present report format is part of the API of webtest and it fulfills its function.

I'm opposed to pure cosmetic changes such as described by (1)

On (2) I agree, this could be done on a backward-compatible way.

On (3), yes, this is sometime useful, and it can also be solved in a backward-compatible way.

The order of the 3 points exposed here isn't related to their significance. (2) or (3) are the motivations and IF compatibility needs to be broken, then (1) should be considered too.

This is not the right place to make such declaration about broken compatibility in your tests. If you have issues, please bring them to the mailing list.

I only partly agree with your sentence "Staying compatible with previous versions should belong to our goals. ". It is incomplete. Otherwise it means that a bad decision can never been fixed. Personally I recognize that I make (many) errors but because I recognize them, I can fix them.

According to the discussion on the mailing list, the WebTest users that adapted the reporting for their own needs are extremely rare. I think this has different reasons (absolutely not sorted by significance):
- the stylesheet provided with WebTest is already quite good
- not so many people are comfortable with XSLT
- the xml report format is over complicated what causes many selectors in XSLT too be over complicated too (see WebTest standard stylesheet)
- WebTest standard stylesheet doesn't provided appropriate extension points allowing to simply extend the stylesheet

(1) and (3) are now implemented and (2) is partially implemented. A new issue could be opened to extend current implementation of (2) when needed, but I close this issue now.