- From: Jason White <jasonw@ariel.ucs.unimelb.EDU.AU>
- Date: Sat, 20 Sep 1997 11:13:12 +1000 (AEST)
- To: WAI HC Working Group <w3c-wai-hc@w3.org>
If I understand Al's model correctly, it seems not to be considerably different from that which has been adopted for HTML 4 and CSS 2. In the original Cougar draft proposal on style sheets, it is stated that to allow for more descriptive media type characterisations, the value of the media type extends until it is delimited by a comma. Thus: media="braille displayed 40, speech" specifies that the associated style or style sheet is applicable to a hypothetical media type of "braille displayed 40" (a 40-cell refreshable braille display) and also to speech synthesizers. Within each of the two media parameters in this example, a genus and species relationship is maintained: the underlying class of medium is given first, followed by a series of qualifiers which, in order of increasing specificity, define the required characteristics of the output device. In the case of a linked style sheet, the first task of the HTML user agent is to decide, as Al has explained, on the basis of the media type, whether to expend HTTP effort to retrieve the file. I agree with Al that the media types could be defined by the style language itself, although I am not sure whether this is possible in a language such as DSSSL (perhaps an expert in this field could provide some enlightenment here). If the media types were defined in the style language, then the algorithm would be roughly as follows: (1) Determine the applicable style language (based on the MIME type specified in the LINK: (E.G. type="text/css"). (2) On the basis of the media types defined in the relevant style language, determine whether the current output device satisfies the criteria specified in the media type. If it does, then retrieve the style sheet. The main drawback of this solution is that it does not accommodate style languages which make no provision for media types. Since HTML is supposed to be independent of all style languages, this limitation could lead to problems in practice. I agree with Al that there is a need for flexibility in the definition of media types and would like to know how an open data base of the kind that he mentiones might work. With respect to CSS, it is clear that the addition of a new media type would require corresponding changes to be made in the style language to support the style properties which are relevant to that medium. This makes it difficult to envisage how an open data base of media types would operate independently of the evolution of CSS itself. Each media type also requires support within conforming applications so that user agents can decide whether the media type matches the apabilities of the current output device. I would appreciate clarification from Al as to whether these comments are on point or whether I have misunderstood his approach completely.
Received on Friday, 19 September 1997 21:13:22 UTC