When I was writing the history of Tp tests for the Talos wiki page I realized that it is time to update Tp.
Good to start with some background first. Tp stands for ‘test pageload’. It is a basic browser test that cycles though a number of web pages. The web pages are culled from the Alexa Top 500. The current version of Tp is Tp4. To create it I grabbed a local copy of as many of the top 500 as I could (some pages always seem to be unreachable). The pages where then narrowed down to the top 100 by following these guidelines:
- Does the page correctly load from a local copy? Are all the images/video present? Does the layout/css still work? Drop if the page is broken in any way.
- Is the page a duplicate? The Alexa Top 500 contains many localized copies of the Google home page and one is enough. Drop if it is a dupe or there is already a similar page in the set.
- Is the page ‘interesting’? Does it contain a font that isn’t covered by another page? Does it have many images? Complicated or large layout? Keep if the page meets any of these criteria.
The question now is if we want to repeat these steps to create Tp5. The main complaint against Tp4 is that the Alexa Top 500 points to the home pages of sites - say the login screen for Facebook or the basic en-US Google page. Yet, these are not things that people are actually using each site for. If we wanted a representative Facebook page it should be someone’s logged in home page; if we wanted a representative Google page it should be a page of search results, or, even better, a Gmail user’s inbox. So, how do we go about doing this? How do we choose the best representation of a given site?
Bug 601798 - create tp5 pageset has been filed to track requirements for the new Tp5 test web page set development. I have some crazy theory that I could post instructions of how to make local copies of web pages and then post a ‘Most Wanted’ list and see if anybody out there is willing to send me content. Anybody up for that? Do you want to have your Gmail inbox be the standard for Mozilla testing? Is that awesome - or terrifying?