I'd rather not. You'd have to make that decision at runtime, in
styleBindAttribute. Something like:
if (type['legacy']) {
node.setAttribute(attribute, value);
} else {
node.acceptAttribute(attribute, type, value);
}
And I think we have learned several times that when we make built-ins behave
differently than user-defined, we pay down the road. Think of how long we have
suffered over <node> is LzNode but <userclass> is userclass.
On 2010-08-25, at 19:29, Max Carlson wrote:
> Couldn't we keep the default encoding for built-in attribute types for
> backward compatibility and only use string encoding for custom types?
>
> Regards,
> Max Carlson
> OpenLaszlo.org
>
> On 8/25/10 2:52 PM, P T Withington wrote:
>> On 2010-08-24, at 19:19, Max Carlson wrote:
>>
>>> LzNode - Use this.acceptAttribute() for style values.
>>
>> This change raises a metaphysical issue.
>>
>> In HTML CSS, every property that is styleable has a fixed type, so the CSS
>> parser knows at parse time, when it sees, e.g.:
>>
>> background-color: blue
>>
>> that `blue` is a special CSS identifier that is to be parsed as an HTML
>> color.
>>
>> In LZX CSS, we allow arbitrary mapping from CSS property to LZX attribute
>> and the type is only known at runtime. Prior to your proposed change, we
>> interpreted CSS identifiers as a call to parse according to the attribute's
>> type, and CSS literal, e.g.:
>>
>>
>> frobozz: "a string";
>> blart: 27;
>>
>> as a call to use the literal as the actual value of the corresponding
>> attribute (no runtime parsing according to type).
>>
>> Your change will break any usage that depended on this old behavior, so I
>> think it needs wider discussion. I'm pretty sure I agree with your change
>> (as it coincides with my proposal to make default values be parsed according
>> to type also). It would simplify our CSS implementation, because it would
>> make, e.g.:
>>
>> background-color: blue;
>> background-color: "blue";
>>
>> equivalent (and the latter just extra typing). You would only need "'s to
>> surround a property value that included whitespace.
>>
>> It seems you need to raise this change with webtop, as they are the primary
>> user of CSS styling.
>>
