- From: François REMY <fremycompany_pub@yahoo.fr>
- Date: Mon, 7 May 2012 22:15:11 +0200
- To: "Tab Atkins Jr." <jackalmage@gmail.com>, "CSS 3 W3C Group" <www-style@w3.org>
(comments inline) > From: Tab Atkins Jr. > Subject: Re: Vendor-prefixes: an idea Happy to see my proposal did please you. It has indeed advantages over the current vendor-prefix situation, but the most difficult part is, like often, overcoming the metastability. Support coming from people inside Google is already a good step. Also, not having a real prefix written inside the property name makes unprefixing seems a lower barrier than it is now. It may be great, since it seems the working group is advancing toward a simpler and shorter path to unprefixing. > On Mon, May 7, 2012 at 7:08 PM, Maciej Stachowiak <mjs@apple.com> wrote: > > One other issue with vendor-bang is that it makes it hard on parsing > > when the !'d version has a distinct syntax from the official, as you > > have to parse the whole property before you know how to parse it. This > > is a solvable problem in principle, but it really makes life easier to > > have a potential syntax switch come at the beginning rather than at the > > end, for example: > > > > foo: !-vendor bar; > > That's not strictly necessary, but it might be necessary given current > parser designs. (If your tokenizer and parser are separate passes, as > soon as the parser sees the property name it can read from the end of > the tokens comprising the property value to look for the vendor-bang, > and then parse the remaining tokens only if it finds the right > vendor-bang.) > > > > Another downside is that the !important syntax is some of the ugliest > > and most confusing syntax in CSS ("not important means important?") and > > it would be a shame to propagate it. > > True. We could potentially prefix them with *any* symbol that > currently always tokenizes as a delimiter. There's a lot of them. Not as many as we would like, however. > > As far as upsides upsides, this proposal makes it seem a little less > > psychologically icky to implement someone else's prefix, since you're > > "ignoring" it rather than "supporting". But the effect > is the same. > > Shane mentioned this, but I'm not sure I agree. While the property is > prefixed, "implementing someone else's prefix" means actively checking > for the other vendor's vendor-bang alongside your own, and considering > the property valid if you find either. However, I can see how it > might feel like a lesser transgression to some people, particularly > since "unprefixing" just means that you stop checking for vendor-bangs > entirely. I really love your interpretation. I didn't think like that myself when I originally proposed this idea, but reading your last paragraph made me think I should have found it myself, because this is really a nice way to see the whole thing. When the property goes "unprefixed", we simply start to ignore vendor tokens. It completely make sense.
Received on Monday, 7 May 2012 20:15:40 UTC