CSS 2.1 Complete: Unblocking Faster Web Standards Progress

Web technologies across the board can make faster and better progress now that the CSS 2.1 specification is a final, completed standard and published as a Recommendation. First, finishing CSS 2.1 unblocks faster progress moving the Web forward with the rest of CSS and other important technologies in the broader W3C HTML5 family.   Even more important, the lessons from CSS 2.1 raise expectations for measuring and improving the actual interoperability of Web standards.

Because CSS 2.1 has been “mostly working” for some time, it’s easy to overlook the importance of this milestone that has been in the making for over ten years. It’s exactly because CSS 2.1 is only mostly working that the milestone is so important. We expect our hard drives, monitors, and keyboards to do much more than mostly work. Finishing CSS 2.1 is a turning point for the Web.

Moving the Web Forward, Faster

Finishing CSS 2.1 provides a stable base for the CSS Working Group to move the Web forward with CSS3. The W3C and Web community recently finished work on the CSS 2.1 specification, completing the final step in the standards process (link) and publishing the CSS 2.1 Recommendation (link, link). The W3C published the final CSS3 Color standard at the same time as the CSS 2.1 standard. CSS3 Selectors and many other technologies (link) can now follow quickly – much more quickly than CSS 2.1 took to complete. Last week, at the working group meeting in Japan (link), the community made progress on a number of new CSS3 features including Microsoft’s grid layout proposal, Adobe’s regions proposal, and the flexible box model. Microsoft also proposed a new module for CSS3 Floats and Positioning.

Finishing a standard is an important part of making the technology site ready for developers and consumers. The problems for consumers and developers of prematurely putting unfinished and unstable technologies into products are covered in a previous post. The related (potentially NSFW) video dramatization of the experience for developers received many comments. The diagram below shows many of the W3C HTML5 specifications under development and where they are in the standards process. Stabilizing and finishing these technologies is important to advancing the interoperable Web’s state of the art.

Diagram of Family of W3C HTML5 Specifications

Professional-grade engineering

CSS2.1 is the most tested and interoperable Web standard today. Its comprehensive test suite sets the bar for all subsequent standards. The community should expect at least this level of testability going in the future.

Testing is an extremely important part of the standards process for developers, consumers, and businesses. It is a crucial engineering pattern for high-quality, robust, industrial-grade standards to achieve real interoperability. The CSS Working Group and community worked hard to deliver this level of quality. Every feature has two or more independent and interoperable implementations, proving that the standard itself can work. The test suite is comprehensive in its coverage of the specification. The community’s work to make sure that all the suite’s tests are valid and that vendors are aware of and can fix the issues uncovered in their browsers is important to make the standard actually work in practice. For any new feature to work in an interoperable way, the community needs it specified unambiguously in a way that everyone can implement it interoperably and test its correctness.

Looking Ahead

The W3C HTML5 specification has reached a significant milestone: Last Call status.

The HTML5 Test Suite is critical for making sure that the same markup – the same HTML, CSS, and script – works the same across different HTML5 implementations. The community will need more tests than the current nearly 30,000 in the suite. Anybody in the community can contribute tests. To learn more about how to help, read the HTML Testing Taskforce Wiki pages and send questions to public-html-testsuite@w3.org. We encourage you to help us finish CSS3 modules as well by creating your own tests and submitting them to the W3C.

CSS 2.1’s completion is a milestone for interoperable and powerful Web standards. We’re excited for interoperable Web standards to keep getting better.

-- Adrian Bateman, Program Manager, Internet Explorer

Comments (22)

  1. Jack says:

    Great to reach this important milestone and excited to see other specs do the same.

  2. Gaurav says:

    When can we expect the next platform preview? It has now been more than eight weeks.

  3. JaiGanesh says:

    Eagerly waiting for the next platform preview. When is it scheduled? Is it planned for the BUILD ?

  4. Mark says:

    @Gaurav They changed the cadence to every 12 weeks to give them more time to provide meaningful changes.

  5. Lea Verou says:

    Since you took the effort of gathering all this data for the chart, why not make it in HTML with clickable links? It would be invaluable.

  6. CvP says:

    If I'm not wrong, 12 weeks means ~5th July?

  7. hAl says:

    IE scores 89.66% of 91.51%  on the CSS2.1 conformance tests.

    link: http://www.w3.org/…/results.html

    That is an impressive pass on 98% of the covered test.

    But why is the IE coverage of the CSS 2.1 testsuite still fairly low. (91.51%)

    Could you please update your coverage on the testsuite?

  8. Tom says:

    Please support WebWorkers and WebSockets.

  9. meni says:

    First, thank you IE team for proving that Microsoft is not all about lock-ins and e^3 (embrace etc.), well done!. Second, these are indeed exciting times where Google and Microsoft and everyone are in agreement on the importance of standards and about moving the web forward. Please make fast(er) progress, things like sound input, webcam support, better JavaScript, etc., so HTML FUDers — I'm looking at you, Silverlight fans, and to a much lesser degree you, Flash developers 🙂 — won't be able to say HTML5 is due in 2022.

  10. Andrew says:

    Sorry for the off topic, but is there any way to drag and drop an URL from IE9's address bar to a Windows Explorer's folder and have a traditional .url file instead of the new .website file?

    (I am not talking about pinning to the taskbar here)


  11. Björn says:

    Huh? I believe that quite some parts of CSS-UI are implemented by Gecko and Webkit, prefixed and not fully standard-compliant but nevertheless.

  12. 6205 says:


  13. FremyCompany says:

    I noticed the font change, too. The fonts are clearer, it should stop the poeple from crying that these are blurry. Anyway, as always, you'll get people who are not happy with the change. The first time I noticed it, I wasn't happy because "it again changed", too. But I must admit it looks better now. What I don't understand is that it's a simple Windows update, and not a service pack or a major IE update.

  14. Adrian Bateman [MSFT] says:

    @hAL – The up-to-date test results are here: test.csswg.org/…/results.html

    This shows IE9 (Trident) passing 98.89% of the tests at 99.99% coverage.

  15. the_dees says:

    @Adrian Bateman [MSFT]:

    Is work going on to improve CSS 2.1 support in IE10?

    There are still many issues in the Trident implementation, some of thsoe aren't even in the CSS Test Suite yet (because the Test Suite was essentially frozen for the last year to stabilize).

    For example: The inherit keyword is implemented incorrectly on almost every property that accepts relative lengths lik em, ch or ex. See Connect Issue 559441: connect.microsoft.com/…/559441

    As always, you can lookup all my tests in an overview: http://www.freewebs.com/…/iepp1

    Some reports are still not reproduced even thoug they have been reported weeks ago.

  16. hAl says:


    Thx for the up-to-date test CSS 2.1 test results.

    Please ask W3C to remove the old pages

  17. Adrian Bateman [MSFT] says:

    @the_dees – Thank you for your continued support in filing these reports. We continue to investigate and prioritise all the issues we receive although sometimes there is a lag before Connect is updated. We do have active bugs in the IE10 database and we will provide updates through Connect when we know more.

  18. Adrian Bateman [MSFT] says:

    @hAl – The page at http://www.w3.org/…/Test has been updated but I believe the old page you linked to was used by the W3C Director to approve the move to Recommendation and my understanding is that they wish to preserve that record.

  19. the_dees says:

    @Adrian Bateman [MSFT]:

    Thanks for your reply. I think today I understand your methods better than when I initially started filing reports. I do not really have a problem with that, because the feeling that you're really working "behind the scenes" improved a lot.

    It's just a little difficult to follow Connect when you're used to more "open" tools like Bugzilla.

  20. Gérard Talbot says:

    @Adrian Bateman [MSFT]

    A lot of what you say is incorrect, inaccurate, exaggerated and not backed up by careful examination of CSS 2.1 test suite [RC6] and [nightly-unstable]. I have tried to edit a message in this blog detailing nicely all this but this IE blog software is very buggy, dysfunctional and unreliable.

    Over 53 testcases have been identified so far as invalid. Many more (=~ 200) are incorrect. Most of them have been authored by Microsoft. If I count all the minor tunings and tweaking, it's well over 500 testcases.

    Gérard Talbot

  21. Gérard Talbot says:

    @Adrian Bateman [MSFT]

    > It is a crucial engineering pattern for high-quality, robust,

    > industrial-grade standards to achieve real interoperability. The CSS

    > Working Group and community worked hard to deliver this level of quality.

    A lot of what you say is not the opinion I have and is not the conclusion I have reached at this point.

    The weakest point of the CSS 2.1 test suite [RC6] is that the perspective of web authors is not met, is not understood. It would take me too much time to explain here.

    The 2nd weakest point of the CSS 2.1 test suite [RC6] is that there is just a) too many lenient testcases, b) too many testcases which can not fail, c) too many testcases which are not really or accurately testing what they were intending to test.

    Btw, this IE blog software is utterly unreliable.

    Gérard Talbot

  22. Gérard Talbot says:

    @Adrian Bateman [MSFT]

    > high-quality, robust, industrial-grade standards to achieve real

    > interoperability.

    The general tone of your post is, in a sense, dangerous, premature and misleading: it conforts and reassures when it should, IMO, be definitely more nuanced and careful. You see, **no one** has actually examine the code of each and all 9000+ testcases of the CSS 2.1 test suite. But from the tip of the iceberg, we can now establish/assess what are the weaknesses of the CSS 2.1 test suite.

    I want to pounce this again: a) too many lenient testcases, b) too many testcases which can not fail, c) too many testcases which are not really or accurately testing what they were intending to test.

    Test suite coverage of CSS 2.1 test suite is excellent … but it is not impeccable, flawless; it is perfectible.

    > The CSS Working Group and community worked hard to

    > deliver this level of quality.

    There is still work to be done. No amount of "completed", "finished", "over with" poetry and "good work" comments-tap-on-the-back in this blog is going to change my mind.

    Fix this IE blog software!

    Gérard Talbot

Skip to main content