Working Group Decision on ISSUE-118 broken-link-types

Question before the Working Group

The current HTML5 draft includes a some definitions of particular link relations that do not exactly match what was defined in previous specifications. In particular, these are the so-called document structure rel values, index, up, first, last, prev and next, and the related or synonymous begin, start, top, contents, toc and end. Some WG members feel this change from past specifications is unwarranted. Others believe that more weight should be given to experience based on authoring practice.

This scope of this decision covers how the set of link relations should be changed in light of compatibility considerations. A section below details what arguments were not considered, a number of which were not considered due to scope reasons.

Uncontested observations

Neither of these were decisive by themselves. There were people who supported each of these three proposals even after taking these facts into consideration.

Summary of Arguments

Three different proposals were submitted, each advocating a different way of dealing with the rel values in question.

Regarding the "Simplify the rel model" proposal

The first proposal advocates the current draft definitions of document structure link relations.

One supporting argument for this particular set of link definitions (and therefore objection to the other proposals) was the claim that existing rel values are "a mess". While it was generally conceded that many rel values are not used with full consistency, the conclusion that they are "a mess" was disputed. There was not a great deal of supporting evidence on either side. Furthermore, if existing usage is poor, that does not necessarily argue for one set of new definitions over another. Thus, this was a weak argument, and a weak objection to the other proposals

Another argument for this change was that document structure rel values are not very widely implemented in browsers. Once again, this does not argue for any specific change; if anything, it calls for removing these rel values entirely, unless they have some use case other than UI in the browser.

A final argument was that WordPress is one of the biggest users of these rel values, so it's worthwhile to be consistent with it. In particular, WordPress uses "index" to mean start page, and no other CMS system cited uses it at all. One counter-argument presented was that WordPress users need to upgrade frequently, so it's more reasonable to invalidate WordPress than other implementations. Furthermore, other CMS systems are more consistent with the historical definitions of document structure rel values. This argument is a wash, because it's not clear why being consistent with some CMS systems is more important than being consistent with others.

Regarding the "Well founded consolidation" proposal

The second proposal argues for several specific changes to link relations. Some link relations are redefined to be more in line with historical usage, and some additional synonyms are added.

One argument for this set of changes is that HTML5's definitions of link types conflict with HTML4, XHTML1, and other previous standards. By itself, this is a relatively weak objection to other proposals - in past decisions, the Working Group has generally prioritized real-world compatibility over matching older versions.

Another raised for this proposal are that HTML5's definitions of link types conflict with some some browser implementations (including Opera). However, it was pointed out that most mainstream browsers do not support these link relations at all; only Opera has UI for them, the UI is hidden by default, and Opera representatives have said it is going to be removed. Thus, this is a weak objection at best to removing link relations or altering their definitions relative to HTML4.

Further arguments in favor of this proposal cited compatibility with use in some CMS systems, notably excluding WordPress; and with use in some specific hand-authored documents (in particular W3C specifications). This is useful concrete evidence. However, it is a relatively weak objection to other proposals. The evidence from CMS systems is mixed, and the set of hand-authored documents does not appear to be a representative sample.

Regarding the "Drop support for rel values" proposal

The final proposal argues for the removal of some relation values, to wit, it suggests removal of index, up, first and last. It was pointed out in survey comments that these relations are already registered in the IANA link relation registry. Presumably, these relations could also be entered in whatever other registry or registries HTML5 adopts for this purpose. In support of this proposal, and against the proposals to retain or alter these relations, a number of arguments were presented.

It was pointed out that support is quite scarce. Document structure rel values are relatively rarely used. Popular Web browsers have no default UI to expose document structure rel values (Opera has UI that is hidden by default), though prev and next are used for some non-UI purposes. Hand-authored pages very rarely use these rel values, though they are often used in template-generated pages. This is a moderately strong argument, and a significant objection to keeping the relation types in HTML5, whether modified or as-is. When an extension mechanism exists, then it does not seem essential to keep marginally important features in the core.

Several objections were presented to this proposal. It was claimed that removal of particular rel values may be ineffective because they are already rarely implemented, and Opera already plans to remove their support. However, this seems like an argument in favor of removal, rather than against.

Second, the SeaMonkey browser as well as browser extensions support document structure link relations, including some that are not defined in HTML5 at all, even in the proposal. But this objection did not explain why the existing registration at IANA and/or possible registrations elsewhere would be inadequate for these use cases. So this was a weak objection.

Finally, it was noted that the link relations being dropped are already defined in the IANA registry and could be defined in future registries; but in the meantime these relations would become invalid, thus invalidating a great deal of content. Since validators do not implement relation checking at all yet and are waiting on a registry solution before doing so, this was also a weak objection.

Overall, the objections to the proposal to remove several relations were the weakest. No objection explained why registration as extensions was insufficient, and such registration would allow more time and freedom to fine-tune the definitions of these relations, independent of the HTML WG.

Decision of the Working Group

Therefore, the HTML Working Group hereby decides to adopt the proposal to drop support for rel="" values based on the lack of interest from implementors and users.

Next Steps

bug 7475 will be reopened and tagged WGDecision, with instructions to the editor to adopt the proposal to drop support for selected rel values. Once the edits are made, the bug and issue will both be closed.

Since the relations to be removed are already registered at the IANA link relation registry, no further action is needed to include them there. WG members are free to register or record these relations elsehwere, as well.

Appealing this Decision

If anyone strongly disagrees with the content of the decision and would like to raise a Formal Objection, they may do so at this time. Formal Objections are reviewed by the Director in consultation with the Team. Ordinarily, Formal Objections are only reviewed as part of a transition request.

Revisiting this Issue

This issue can be reopened if new information come up. Examples of possible relevant new information include:

Additionally, bug reports on the definitions of relation types that remain in the spec continue to be welcome.

Arguments not considered