Re: Proposition to change the prefixing policy

> Again, I state that authors should use the newest draft (or the unprefixed
> version once it is a CR). The older drafts only should be supported to
> prevent older websites from breaking. Older drafts can be used if the 
> newer
> draft is not fully supported by all browsers. Everyone can look on a still
> to be made website to see which draft is supported.
>
> it should look a bit like http://caniuse.com/, but then with each draft 
> vertically and the browsers horizontally.
>
> This way of prefixing could probably not be implemented until the next 
> browserversion comes out. When it does it supports the older drafts, so it 
> isn't needed to show at which browserversion it is supported, because all 
> drafts would be supported by the newest browser.

The main issue remains: blog posts, printed books and similar resources are 
static snapshots of a moment in the spec history. Most people use them, but 
few will "live update" to reflect changes in support, modify recommanded 
prefix. That means the code "copy-pasted" from those sources will not 
necessarily be in sync with what you would like to; old prefixes won't die 
easily. Also, this solution would lead to an important amount of 
per-property prefixes. It'll hurt parser performance, at some point.

I'm pretty sure an intermediary approach exists between "all vendor-based" 
and "all version-based".

Anyway, the best solution remains no prefix at all, as soon as possible. 

Received on Tuesday, 8 May 2012 11:30:38 UTC