Re: [lots] -webkit prefixed properties and values

On Wed, Dec 9, 2015 at 2:15 PM, Dean Jackson <dino@apple.com> wrote:
>> On 10 Dec 2015, at 08:19, Tab Atkins Jr. <jackalmage@gmail.com> wrote:
>> Per the Compat Spec <https://compat.spec.whatwg.org/>, there's a
>> decent-sized list of CSS at-rules, values, and properties that need to
>> be supported with a -webkit- prefix in order to be web-compatible:
>> <https://compat.spec.whatwg.org/#css-compat-section>.
>>
>> Since implementors have to support these in order to realistically
>> support web content, they should be listed alongside the features in
>> the relevant specs (rather than sidelined into an easy-to-miss errata
>> document like they currently are).
>
> I disagree. The existing implementors (obviously) know about
> these properties. New implementors are unlikely to start from
> scratch. And even if they do, the number of new implementors
> that appear each year can be rounded to about zero.

As Ms2ger responded, new implementations *do* show up, and when they
do, this knowledge is *not* obvious or ingrained.  Servo is a new
project *within Mozilla* and they've had to do a lot of work to
transfer over some of this "implicit knowledge".

> Honestly, I don’t think it’s worth advertising these properties
> any more than they currently are. The sooner we stop talking
> about them, the sooner we can remove support (even if that
> is many years away).

It would be listed as an "implementors MUST, authors MUST NOT" sort of
thing.  Pretending they don't exist isn't helping existing
implementors.

>> I'm planning to do this for all the specs I control.  Would others
>> please do the same?  The specs in question are:
>>
>> * Images
>
> I assume here you’d have to describe the legacy gradient
> syntax that WebKit implemented before the specification
> changed?

According to the Compat spec, yes.  Edge and Nightly Firefox have both
implemented that. Servo would have to as well, as would any other new
implementor.

> This is another example of why I think they
> shouldn’t be in the primary specification: I don’t want any
> Web authors discovering them. The specification should
> only talk about the correct way of doing things. The compatibility
> specification seems like the right place for old/incorrect
> or deprecated stuff.

This is not our general policy for errata and legacy stuff.  We
usually put them in an appendix with a strong warning for authors not
to use them, like I'm proposing.

~TJ

Received on Wednesday, 9 December 2015 23:18:28 UTC