- From: Al Gilman <Alfred.S.Gilman@IEEE.org>
- Date: Fri, 20 Jul 2007 08:27:35 -0400
- To: www-tag@w3.org
In my recent post I tried to establish the case that semantics for @class, linked through @profile (either page-global in HTML4 or more locally in HTML5), is backward-compatible. http://lists.w3.org/Archives/Public/www-tag/2007Jul/0079.html Now I'd like to take on a slightly broader horizon. In his initial post, Dan muddied the waters by leaping to a draconian step: taking loosey-goose tokens under spec control. That's not implied by the milder step of establishing a preferred way of exposing interpretation guidance for the author-controlled tokens in attributes that allow such. There are cases where concepts need to be migrated from crowd-sourced standing to spec-sourced standing. The justification would have to be an exigent need, comparable to the legitimate use of "taking by eminenet domain." But in this case new signs for the newly-controlled senses *are* a better way to go. Step 1: Make it clear that it's cool within @profile (or <link rel="dcterms:conformsTo" ...> or whatever) to introduce references to data dictionary documents explaining @class and @rel/rev values in addition to values of meta.name. Step 2: There will take some time to see if we get any takeup of data dictionary practice. Step 3: There will take some further time for a shakeout in data dictionary languages. Step 4: shortnames in the domain of distributed coinage and glossarization as above will, through viral propagation and genetic competition, develop some common usage. HTML could, as an efficiency measure, introduce, by spec provision, a dictionary of the most stable of these into the collection of data dictionaries cited in @profile *at the low precedence end of the pecking order* comparable to the 'initial' values set by language default in CSS, that lose out to all other defaults and values set more specifically on a case by case basis. Step 5: In some cases, meanings associated with terms formerly crowd-sourced will have processing or trust dependencies that make an irrefutable case for bringing them under control. The problems vexing the Web Security Context WG at present suggest one domain of practice where this could happen. In these cases, as Mark argued, there is reason to introduce new signs for the newly-controlled senses. If there's no version checking, then there should probably be new markup, amenable to discrimination with existing selectors. But the simple fact that people aren't checking the data dictionaries for terms that look familiar, that's not sufficient justification for the more radical step. Best to leave the semantics that processors don't honor in the crowd-sourced bucket; rather than be subjected to the ignominy that nobody is paying attention to the format spec proper, all over again, just the same. The spec can, for established usage, play nice and beneficially in the crowd-sourced puddle, if it takes a humble enough seat at table. Al
Received on Friday, 20 July 2007 12:27:47 UTC