- From: Richard Ishida <ishida@w3.org>
- Date: Fri, 29 May 2015 19:21:52 +0100
- To: Koji Ishii <kojiishi@gmail.com>, Florian Rivoal <florian@rivoal.net>
- CC: "Phillips, Addison" <addison@lab126.com>, "cowan@ccil.org" <cowan@ccil.org>, "Elizabeth J. Pyatt" <ejp10@psu.edu>, www International <www-international@w3.org>
On 29/05/2015 18:35, Koji Ishii wrote: > On May 28, 2015, at 06:02, Florian Rivoal <florian@rivoal.net> wrote: >> >>> On 27 May 2015, at 18:34, Richard Ishida <ishida@w3.org> wrote: >>> >>> On 27/05/2015 10:09, Koji Ishii wrote: >>>> Thank you for the great article, Richard. >>>> >>>> A couple of comments: >>>> >>>> 1. It’s probably better to put unprefixed property at the end. >>> >>> yep, done. >> >> Actually, I'd rather you didn't. This is a lost battle on most properties given how often this is done, but this approach has problematic consequences, and unprefixed writing-mode usage so far might still be sufficiently rare that we're not yet stuck. >> >> TL;DR: If too many authors follow this advice, it may get in the way of desirable spec improvements and fixes. >> >> Full version: >> >> Browsers don't support the propoerties unprefixed because while the spec is fairly stable, it's not stable enough yet to be considered final, and we may still want to change things. >> >> By writing the unprefixed variant as well, desipite the lack of existance of unprefixed implementation, you are making the bet that there will be no change between now and final implementations, and that whenever they come around, final implementations will interpret your unprefixed code the same way as they now interpret you prefixed code. >> >> But it is precisely because we're not sure that this will be the case that prefixes exist in the first place, and for writing modes we have actually discussed potentially breaking changes at the last F2F. >> >> If enough people use the unprefixed properties before unprefixed implementations exist, to avoid breaking existing content browser vendors may be prevented to make a behavior change that would otherwise be desirable, defeating the point of having prefixes. >> >> Had content predating the desirable spec change only used prefixed properties, the browser vendors could have kept their existing behavior for the prefixed version, while implementing the new and better behavior for the unprefixed implementation. >> >> This "best practice" problematic enough that nowadays most browsers, instead of shipping work-in-progress features under a prefix like they used to, actually don't ship them at all in production builds until the spec is final. >> >> - Florian > > I see, thank you for the full version ;-) > > I agree that flags are better than prefix, but we will need to deal with until all prefixed properties are gone. > > For that, I was blindly following what people say, and echoed to other people, but your explanation makes more sense to me. > > /koji The un-prefixed property name is there because Firefox doesn't support -moz-writing-mode, etc. It only supports un-prefixed property names. That said, it only supports anything at all in nightlies (unless you explicitly turn it on in the released version, of course). (Note, btw, that IE has supported un-prefixed property name for writing-mode for many years, with a very different model. So actually things are already pretty broken.) I could add some wording to say more clearly: here is what the spec currently says - and here's what currently works in browsers. ri
Received on Friday, 29 May 2015 18:22:04 UTC