- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Fri, 18 Nov 2011 09:51:56 -0800
- To: "Marat Tanalin | tanalin.com" <mtanalin@yandex.ru>
- Cc: Philip Jägenstedt <philipj@opera.com>, Tantek Çelik <tantek@cs.stanford.edu>, Sam Ruby <rubys@intertwingly.net>, public-html@w3.org
On Fri, Nov 18, 2011 at 9:22 AM, Marat Tanalin | tanalin.com <mtanalin@yandex.ru> wrote: > 18.11.2011, 19:59, "Philip Jägenstedt" <philipj@opera.com>: >> We cannot provide equivalent markup for itempropvalue unless we know how >> itempropvalue is intended to work: >> >> 1. When itempropvalue appears on an URL property elements [1], is it (1) >> ignored (2) resolved as a URL or (3) a plain text property? > > For URL property elements, @itempropvalue attribute should probably be ignored since such elements have their own dedicated attributes to store a value. That's even worse, imo. @itemscope already "wins" over special attributes when they're specified together (in other words, in <a itemprop=foo href=bar itemscope>, the value of "foo" is the nested item, not the url). If @itempropvalue instead lost to the special attributes, that would be all kinds of confusing. >> 2. When itempropvalue appears together with itemscope, is it (1) ignored >> or (2) a plain text property? > > @itempropvalue should be treated according to @itemtype. If @itemtype defines that item value should consist of several values specified by child elements, then @itempropvalue for element with @itemscope attribute should probably be ignored. The Microdata extraction algorithm doesn't pay attention to the itemtype when extracting. It *can't*, because there's no schema for indicating such a structure in a machine-readable way, there's no way to link the itemtype to a schema in a machine-followable way, we know from long experience that adding such a mechanism (see DOCTYPEs) is an invitation to DDOS the provider, and finally, if the resource is ever unavailable, either temporarily (site overwhelmed) or permanently (___domain registration expired), the extraction algorithm would have no way to know what to do. Now, *after* the data's been extracted, a type-aware consumer can interpret the data however they like. They've got the schema built-in, so none of the above problems affect them. ~TJ
Received on Friday, 18 November 2011 17:52:48 UTC