- From: Didier PH Martin <martind@netfolder.com>
- Date: Sat, 28 Sep 2002 10:54:43 -0400
- To: "'Eric van der Vlist'" <vdv@dyomedea.com>, <www-tag@w3.org>
Hi Eric, [Eric proposal to explore using CSS for Hlink to xlink mapping...] Didier replies: Ok this is a working hypothesis and this avenue needs to more explored. Here is another one: Premises: ---------- a) browsers support XSLT 1.0 (actually, Mozilla and IE, probably Opera in a near future - Small devices browsers like mobile phones do not support it, only XHTML 1.0 + CSS. b) XHTML 2.0 proposes more semantic elements and less rendition elements. c) most of the browsers support a DHMTL visual model. In fact HTML or XHTML is used to encode a hierarchy of visual objects. There is a one to one equivalency of most of the HTML (or XHTML 1.0) elements and a visual element (with the exception of some non visual elements) Proposal: --------- Since, XSLT transformations are possible either on the server side or on the client side and since the number of XSLT enabled browsers is growing, then a clear separation of the rendering and the semantics is made more and more possible. To encourage this view/model pattern, XHTML may evolve toward being a language with more semantics. IN fact XHTML 2.0 shown some sign that this is the case. An XHTML document could be transformed for rendition with an XSLT style sheet or with a CSS style sheet. This transformation establish a one to one mapping of certain elements/attributes into visual objects. However, with the actual state of the arts, it is very difficult to encode images as semantics, probably at the meta level but hardly at other levels. Thus, there is a very high probability that visual object such as SVG or MathML objects are to be embedded into XHTML documents. Two choices are offered a) the visual objects are defined into the XHTML document b) the visual objects are external to the document and referred with a link. In case (1): the visual object is embedded into the XHTML document. The XHTML document can be made more coherent and logically consistent if all links are expressed in the same way. For instance, that an SVG, mathML or XHTML links share the same characteristics and syntax. An <a> element in SVG is also identical in XHTML and if both share the same characteristics, then the overall document is more readable, understandable, easier to maintain. It has been proven in software psychology - mainly through the work of Ben Shneiderman (1) - that coherence is an important factor to reduce errors, reduce the learning curve. Moreover, coherency may cause a network factor and induce other domain language authors to follow the recipe. In case (2): the visual object is external and referred with a link. If XHTML takes the path of having a single element for anything to be embedded in the final rendition (i.e. the <object> element). Then this object could be expressed as an xlink extended. This may require an upgrade of the xlink specs. However, this update shouldn't break the other language using xlink 1.0 already. We are fortunate that these modification are possible without causing a breakdown to other domain languages using Xlink 1.0 (by using some defaults premises like arc direction, etc.). In a previous post, it has been shown that new possibilities are brought if XHTML uses xlink. In previous post, we also notices that Xlink needs some specs upgrades. The capacity to use an <a> element (as a simple link) or to overload it to be an xlink extended element brings something more. Up to now, it has not been proved that not using xlink or using Hlink brings new capabilities to XHTML. To the contrary, it just add more processing or more information to add to an XHTML document. Moreover, it causes a major breakdown for an XHTML document to be processed on the client side with an XSLT style sheet. It does not reduce the complexity nor it brings us more than we have today. Just more constraints and potential added costs. Cheers Didier PH Martin
Received on Saturday, 28 September 2002 10:54:50 UTC