- From: Robin Berjon <robin.berjon@expway.fr>
- Date: Mon, 30 Aug 2004 11:48:29 +0200
- To: Cameron McCormack <cam-www-svg@aka.mcc.id.au>
- Cc: www-svg@w3.org
Cameron McCormack wrote: > I was thinking that if you defined the trait then you would be able to > then style your element with that CSS property. But that isn't > possible? There's a number of issues in there. The fact is that traits are imho pretty much still in flux in terms of how far they reach and what exactly are the limits of their functionality. Initially you are right, and they would have made new style properties defined for some custom elements (even though it's unclear to me that that had any impact as the said properties could be made available as unknown properties anyway). I can't speak for the WG here but it might be better for everyone's sake if only XML attributes were supported, at least in this version. There is enough disagreement as to how CSS and SVG marry that we might not want to add to it. > That was my worry, that there would be no CSS syntax to use > for specifying the matrix value. If it is only for animation then I > guess there is no problem. I would assume that where applied to CSS it would use whatever syntax SVG would use for that type. This would effectively create a new function type in CSS: matrix(a, b, c, d, e, f). That's ok, except that in SVG the following is also a valid matrix: matrix(a b c d e f) whereas in CSS function arguments aren't space separated. This is one (amongst quite a few) situations in which the side-effects from traitDef influencing CSS are imho cause of more trouble than they bring value. In the interest of keeping SVG 1.2 implementable, I think that traitDef should not influence CSS. There are use cases for making arbitrary CSS available to arbitrary markup, for instance to make all your buttons blue -- in fact CSS all of a sudden becomes a lot more useful as one starts using a vocabulary that isn't SVG -- but I think that those can be addressed fully if the said CSS properties are simply made available to those elements as unknown properties: you would ask the value of for instance the property 'button-fill' (a terrible example, but let's pretend) and you'd get back a CSSValue the cssValueType of which would be CSS_CUSTOM. In order to use it you would simply parse its cssText yourself. To me this seems sufficient, and would alleviate implementation weight while at the same time avoiding a whole mess of cross-spec interaction. All traits would be of type XML. Bridging that with SMIL would work easily (the code is mostly there), and quite likely be sufficient for most uses. > Is there any advantage to being able to specify whether the trait is of > type XML or CSS? Is it that CSS traits will be inherited and XML traits > not? CSS traits would be specifiable in a stylesheet. Their default inheritance behaviour would need to be specified (it currently isn't), which is another of those thorns. > Maybe I should just wait for sXBL to come out. :) I misled you by pointing at sXBL, I'd forgotten that traitDef had been moved out of that area and into pure-SVG 1.2 stuff. sXBL is indeed very forthcoming, but traitDef isn't in there. -- Robin Berjon
Received on Monday, 30 August 2004 09:49:00 UTC