- From: Maciej Stachowiak <mjs@apple.com>
- Date: Tue, 18 Nov 2008 16:02:02 -0800
- To: elharo@metalab.unc.edu
- Cc: Boris Zbarsky <bzbarsky@MIT.EDU>, Ian Hickson <ian@hixie.ch>, "Henry S. Thompson" <ht@inf.ed.ac.uk>, public-html <public-html@w3.org>, www-tag@w3.org
On Nov 18, 2008, at 8:13 AM, Elliotte Harold wrote: > Maciej Stachowiak wrote: > >> (Of course, many of the conformance criteria relevant to mainstream >> browsers won't apply to other kinds of software. Fortunately, HTML5 >> defines a number of conformance classes: <http://www.w3.org/html/wg/html5/#conformance-requirements >> >. The classes "data mining tool" and "markup generator" cover >> between them any tool that takes HTML as input or output, and are >> free of browser-specific concerns.) > > Sadly this isn't true. For instance, in data mining, "Applications > and tools that process HTML and XHTML documents for reasons other > than to either render the documents or check them for conformance > should act in accordance to the semantics of the documents that they > process." > > This statement indicates an underlying belief that semantics are > interoperable and that a document can in fact be said to have a > particular set of semantics which can be shared and transferred > between independent parties. > > Of course, I expect most data mining tools to more or less ignore > this statement and infer whatever set of semantics they need from > the syntax. RFC2119 SHOULD means the requirement does not apply if there is any significant reason to do otherwise. For a search engine for instance, the fact that better results can be produced by often ignoring defined semantics would be a good enough reason. - Maciej
Received on Wednesday, 19 November 2008 00:02:48 UTC