Anne van Kesteren

HTML5 timeline

Clearly we suck at marketing. Publishing an A List Apart article with an honest timeline is not the way to go. (Though publishing an article there clearly is.) Firefox broke its implementation of the HTML5 canvas element and guess what, the world noticed. Actual websites started breaking because they relied on the canvas functionality to work. The point is that we expect implementations of HTML5 to happen way before the fifteen year mark. In fact, the fifteen year mark includes having all features at least implemented in two different (shipping) products in the same way with the additional requirement that they have a decent amount of market penetration. This means that when the specification finally makes it to W3C Recommendation it has already proven itself. (Or it does not get implemented and dies. This seems unlikely though given the involvement of all the key players in the browser market.)

As a comparison point we sometimes use CSS2. That effort started ten years ago or so and still has not reached the level of quality we want from HTML5. And also not the level of quality that the CSS Working Group demands for that matter. CSS 2.1 is cleaning up the CSS2 specification and will not become a W3C Recommendation until there are two complete interoperable implementations.

It is not that we finish writing the specification in fifteen years and browsers will start implementing when we’re done writing. It is that in fifteen years we expect HTML5 to be a fully integrated part of the Web with rock solid implementations you can rely on. And as with CSS2, parts of the specification are already usable.


  1. Call me old fashioned, but what HTML5 needs is a bit less "marketing" and a bit more test suites. Maybe it will only take us 12 years to get there.

    Posted by Sam Ruby at

  2. I think we need both. I don’t really mean marketing by the way, just means to get attention from our target audience so we can incorperate their feedback and ensure that important arguments don’t get lost. As for test suites, as features are being implemented people have been developing them. There are quite extensive test suites available for canvas, Web Forms 2.0, and HTML parsing for instance.

    Posted by Anne van Kesteren at

  3. The parts that are already usable are the parts that overlap fully with (X)HTML.

    Posted by Joe Clark at

  4. Would this extensive test suite for canvas have caught the bug that was noticed by the world? Is the Mozilla team unaware of this test suite?

    Why does the recently created FAQ entry say There is no real test suite?

    Where can this test suite be found?

    Posted by Sam Ruby at

  5. See <canvas> tests. Philip recently took some time to make them ready for Mozilla after the Firefox fiasco: Bug 407049. As far as the FAQ entry goes, nothing is really official yet. Test suites are currently ad hoc. Once implementations get more mature we can invest some time in making the tests for that particlar part of the specification more official.

    Posted by Anne van Kesteren at

  6. I meant making them ready for the Mozilla regression testing system. Also, it would indeed have caught the issues with Firefox

    Posted by Anne van Kesteren at

  7. Excellent!

    Posted by Sam Ruby at

  8. Speaking of HTML 5, it would be interesting to hear your take on the Vorbis/Theora issue.

    Apparently, Nokia and Apple have succeeded in kicking the Ogg word from the spec?

    Posted by David Naylor at

  9. If it wasn't for people (like Anne) pushing the envelope in the first place -- with new and, dare I say it, experimental elements in their sites -- we could probably all safely add another 10 years to those timelines.

    I mean:

    ul#nav:after {
      margin:20px 0pt 0pt -15px;

    Who else would have ever dreamt up something this innovative?!

    Keep up the great work. It always causes me pleasure -- if a little associated pain (working it out) as well. ;)

    Posted by iNSiPiD at

  10. it would indeed have caught the issues with Firefox

    There is still nothing automatic to stop the same issue occurring with Firefox, since the regression tests were only added to the trunk (i.e. Firefox 3) – most of the testing infrastructure is quite new, and I haven't noticed any interest in backporting it.

    At least Firefox 3 does have these tests, so it will help in the future, though it'd be nicer if the Mozilla developers had taken the initiative in setting up the tests themselves. The impression I get is that Mozilla has reached the stage of accepting that testing is important, but not the stage of deciding that it should be required for all new code.

    Posted by Philip Taylor at