Re: [CSS21] Lack of version control for content

Chris Lilley wrote:
> On Thursday, August 25, 2005, 9:24:39 PM, fantasai wrote:
> 
>>>As an example, @version 2.1; would be one method. Unknown at-rules are
>>>ignored by compliant parsers.
> 
> f> This issue was discussed at the previous F2F and I believe it was
> f> concluded that CSS does not need to embed version information.
> 
> I would e interested in a pointer to that discussion, offlist oif
> required.

Can't seem to find any.. I'll dig around in my IRC logs. The discussion
happened after right we'd officially closed the meeting, iirc, so I
don't know if it was officially minuted or not.

> f> Instead, the validator software should allow for multiple profiles,
> f> including tailored ones based on the actual state of relevant
> f> implementations
> 
> you mean, @aimed-at WinIE5;  or something else ?

I mean, you write your style sheets as you do now, with no version
or profile comments, and when you want to validate it you tell the
which profile to validate against. This could be done by selecting
an item on the validation form, or by accessing a separate validator
URL for each profile, or by setting your config files if you're
validating locally -- it doesn't matter, that's a UI issue. If you're
authoring CSS for a web site, the importance of validation isn't
about restricting yourself to a particular version, it's about making
sure your syntax is correct and making sure you aren't using features
that you know will break in the implementations you intend to support.
There's no real reason to tie your style sheets down to a specific
version: you just validate against the profile you want to restrict
yourself to. As that changes with advances in implementations, you
change your validation method, not all your CSS files.

Ideally the validator will use some kind of easy syntax for defining
profiles. It should offer profiles for CSS levels 1-3, other CSS
Profiles such as the Mobile Profile, and custom profiles. Given such
a capability, the web authoring community could tailor and maintain a
few config files representing CSS features that are implemented well
enough to use widely on the web, regardless of whether the features
were defined in CSS1, CSS2, or CSS3. There are some CSS3 features that
are already more interoperably implemented than some CSS2 features,
so this kind of validation would be much more useful to authors than
validating against a CSS specification level profile.

> f> rather than less relevant state of version numbers.
> 
> So, are you saying that
> 
> - the introduction of CSS 2.1 does not add any
> particular value since all implementations do different things
> - validation is useless for CSS (please tell the validator developers if
> so)
> - something else?

See above. The value of CSS2.1 is that it fixes many problems in
CSS2, clarifies many ambiguities, and brings the spec in line with
implementations and implementors expectations. By trimming out some
completely unimplemented features it also defines a more realistic
target for implementors. CSS2.1 goes a long way towards getting us
interoperability and towards providing a more solid base for CSS3.

~fantasai

Received on Friday, 26 August 2005 15:46:59 UTC