Hixie's Natural Log

2003-06-22 22:38 UTC Half a weekend

I saw GwieF and Phil (and Claire) today. We went to see 2 Fast 2 Furious, a fun movie. Amusingly few cars actually blow up in this film, which makes an interesting contrast with the usual fare, where if any two objects collide, a fireball is guarenteed, even if there are no flammable materials within a 10km radius. In fact, as GwieF pointed out, the accidents in this film were often underpowered, with cars coming to a calm standstill after hitting a lamp post at 180kph.

Earlier today I attached a patch to bug 38370. It basically replaces Mozilla's <hr> implementation with "pure" CSS, fixing tests like this. It also fixes some existing layout and compatability bugs: it removes the HR-quirk that caused several layout and style problems, it adds support for the (non-standard) color attribute, and it solves a long standing problem whereby Mozilla couldn't draw rules smaller than 3 pixels tall.

I don't do much in the way of Mozilla bug fixing, mainly because each time I look at the code I get frightened stiff. I find it amazing that code this complicated works as well as it does. Thankfully, people like David are working hard to simplify it, so maybe one day I'll be able to look at, e.g., the line layout code and not get a headache.

In the meantime, Hyatt has been rambling on about standards compliance charts. I have to agree with him on most of his points, except one: test suites should never ever be scored by percentages. Too many times have I heard the term "100% compliance" (meaning "we didn't spot the 5 bugs in this simple test suite, and so we reckon that our implementation is perfect"). No, much better is to score implementations, giving points for tests that pass, and subtracting points for tests that show bugs (and not giving any points if there is no implementation — a bug is much worse than having no support). In my opinion, that gives you a much better idea of how reliable implementations are.