- From: Orion Adrian <orion.adrian@gmail.com>
- Date: Mon, 13 Jun 2005 08:59:25 -0400
- To: www-html@w3.org
This is something I had thought about some time ago, but decided to post my thoughts here after reading the heading (<h>/<hx>) dicussion. HTML does several things that perhaps it shouldn't or should be further modified. 1) It attaches metadata to the document. Personally I'm of the mind that HTML shouldn't do this since it should be the responsibility of the OS to store this data in a general sense as part of the file descriptor. Each document having it's own metadata structure just makes it that much harder for the OS to work like that. However a generic container for metadata that is a transfer protocol could be useful. 2) It organizes documents into a container. Each top-level section in a document is a sub-document. Each has it's own title and while the sub-documents are linked, the contents of each are not continuations of each other. Each sub-document could live on it's own, though in this case it would probably detract from our usage of the documents. Though with this in mind I would decrecate <body> for multiple sections. More on this below. 3) It marks up outline structure. This is something it actually does very well. Well as long as <hn> goes away. We should also get rid of <div>, but that's another problem entirely. Otherwise it does its job very well. 4) It marks up content for semantic classification. This is doesn't do so well. It does an adequate job, but seeks to apply tree structure to data that doesn't have a tree structure. It rallie against empty, replaced, semantic elements which are a perfectly valid case. While entities could be used, they aren't necessarily beneficial. Also the semantic classes in HTML seem more programmer oriented than other traditional bodies of work (e.g. <code>, <blockcode>). I'm guessing that if you could find someone in the literary world, they would ask for a few more especially in the references section. Maybe subtitle and the like. Content classification I believe should be more plug-and-play. This should be sub-classed out. To make some of these changes, I would do the following. 1) Take out linking structures and turn them into their own language but simplify them (XLink goes overboard). A simply @href attribute would suffice. 2) The metadata part of a document should be it's own specification used by anyone and should be limited to a transmission protocol so that OSes can manage their own metadata. 3) Deprecate <body>, replacing it with <section>, <table>, <object> or <list> (or other top-level elements from other specs (svg, spreadsheet, slides, etc). 4) Create an element called <package> used to place multiple top-level elements into it. 5) Fix namespaces (just thought I'd throw this in everywhere I get the chance). Orion Adrian
Received on Monday, 13 June 2005 12:59:33 UTC