HTML Ruby PR Escalation to WHATWG SG

Hello WHATWG SG,

This message concerns “ruby”, which is a form of interlinear annotations commonly used in East Asia, primarily (but not exclusively) for phonetic assistance. It is an important typographic feature as well as a critical accessibility feature for languages such as Japanese, where it is notably relied on by virtually all language learners, including by all school-age children.

I'm writing specifically to escalate the matter of the rejected PR to synchronize the W3C and WHATWG HTML ruby specifications.

Aside from substantially rewriting the prose and examples, the pull request does three things:

adds a row-primary ruby markup pattern using RB
removes Hixie's double-rt double-sided ruby feature that nobody implements
defines RTC to provide double-sided ruby under the row-primary model

This PR resolves what is, afaik, the only major intentional difference between the WHATWG and W3C HTML specs (prior to the MOU), incorporating the broadly consensus-backed W3C extended ruby markup model into the WHATWG spec.

We have two implementations of the first (RB) feature: Mozilla Firefox and Amazon Kindle. (But I am told that Kindle does not count as far as WHATWG is concerned, since it is not a "Web browser".) The second (RTC) feature is currently only implemented by Mozilla afaik (and is somewhat less critical).

The background on this PR starts more than 10 years ago; to understand it probably the best place to start is this blog post: https://fantasai.inkedblade.net/weblog/2011/ruby/

For more examples and explanations there's also the content of the HTML PR, which is posted here: https://html.rivoal.net/multipage/text-level-semantics.html#the-ruby-element

These two should explain the use cases and requirements, and the resulting technical design, of the extended markup model.

From a historical perspective, fantasai's post kicked off two efforts in coordination with the i18nwg:

https://www.w3.org/TR/html-ruby-extensions/ in the HTMLWG This eventually got incorporated into W3C HTML5, but, due to Hixie's objections to adding anything more complicated than what he personally thought was adequate, never made it into WHATWG HTML.
https://www.w3.org/TR/css-ruby-1/ in the CSSWG The modern CSS Ruby Layout spec, which handles all the markup patterns described in the blog post.

Thanks to Koji Ishii, Robin Berjon, Henri Sivonen, and the i18nwg's efforts, all browser engines implemented the relevant parsing rules for the full set of ruby elements awhile back, and this has been incorporated into the HTML parsing algorithm across all specs and implementations.

Subsequently the Ruby Markup Extension spec was retired as a W3C NOTE and its contents incorporated into the W3C HTML REC.

Xidorn Quan (@upsuper) then implemented the new CSS Ruby layout model in Gecko, and used it to implement layout support for the full HTML ruby extensions spec in Firefox. He later filed an issue to have WHATWG sync with W3C's HTML spec, which then stalled.

At the 2018 TPAC the W3C i18nwg and WHATWG editors discussed the situation of HTML ruby markup, as it was a major technical issue that the W3C community found unsatisfying in the WHATWG specification. The i18nwg promised to provide a PR, and the WHATWG would merge the PR provided Firefox's CSS implementation and a second layout implementation commitment.

This PR was provided in March 2020, at which point the WHATWG editors clarified that the second implementation must be "a second implementation approved for shipping" of support for layout of the extended markup, and cannot be anything but WebKit or Blink because WHATWG policy only considers web browsers to be implementations. See https://github.com/whatwg/html/issues/1771#issuecomment-796976770

Thus currently we find ourselves in an awkward position:

The HTML ruby extension spec and HTML5 REC are retired as well as somewhat outdated.
The editors of the current HTML spec are unwilling to merge the PR because there is no second shipping Web browser engine implementation
Because no maintained specification exists, we have implementers (such as Amazon) working off of old, unmaintained specs, which creates a poor environment for interop and cooperation.

There are, afaict, two ways out of this situation:

WHATWG merges the PR without a second browser implementation
- Benefits of this approach are that the HTML spec stays in one place.
- Detriments are that the WHATWG community seems difficult to work with on this matter, which could impede further improvements.
W3C publishes a Ruby Extension Recommendation
- Benefits of this approach are that the specification is worked on where the most relevant expertise already exists (i18n, a11y, css).
- Detriments are that this portion of the HTML spec will be maintained outside the otherwise-complete WHATWG living standard.
Note: In this second case, we would provide a PR to update the WHATWG specification to remove the features that were never implemented and to ensure that, in both technical and editorial respects, the specifications are synchronized and appropriately cross-referenced to each other such that the WHATWG specification clearly defines a semantically compatible subset of the W3C extended ruby model.

The question I am putting before the WHATWG Steering Group is, which of these two approaches would it prefer to take?

~fantasai

The Steering Group has reviewed this escalation and members of the SG are amenable to the prospect of having a Ruby Extension specification published by the W3C according to option 2 noted above. The SG is interested in further validating this decision with its HTML community and has filed https://github.com/whatwg/html/issues/7405 to solicit any feedback and concerns. Pending substantial negative feedback (and we should give about a month for folks to respond given we are entering into the end-of-year holiday season), we believe an updated Ruby Extension spec, kept in sync as much as possible with the current Ruby prose in the HTML specification, is the best way forward. We appreciate your interest in preparing a document that we all agree can eventually be integrated back into HTML when browser implementations of the extension are ready or near-term committed.

We understand that the W3C will specify the extended HTML Ruby markup described in PR#6478 as a REC track document. We believe this is within the bounds of our MoU with the W3C, and while it is not the ideal outcome (having Ruby documented in two places), we understand that it serves the goals for the W3C's constituencies and existing implementations to have an easily referencable document as well as honoring the strict requirements of the WHATWG that the living standard not contain new additions that don't have multi-browser-engine implementer commitments.
We understand that the SOTD section of the W3C Ruby Extension spec will state our joint intention that the extended ruby markup definitions be fully integrated into the WHATWG HTML specification once they meet WHATWG's inclusion criteria under section 10.2 of the 2019 MOU.
We understand that W3C editors will offer pull requests to keep the HTML spec in sync, both technically and editorially, for the subset of features defined in both specs (goal is to reduce the deltas such that the only differences will be the addition of the two new elements <rb> and <rtc>); we expect WHATWG HTML editors to work with the W3C Ruby Extension spec authors to merge those pull requests in support of that goal.

By mid-January 2022, barring any substantial negative feedback from the HTML community, this plan shall be committed and documented in the W3C/WHATWG Coordination repository.

whatwg / sg

HTML Ruby PR Escalation to WHATWG SG #184