Re: Draft [URL] reference update to informative text

On 10/09/2014 04:55 PM, Yehuda Katz wrote:
> On Thu, Oct 9, 2014 at 1:32 PM, Sam Ruby <rubys@intertwingly.net
> <mailto:rubys@intertwingly.net>> wrote:
>
>     On 10/09/2014 03:44 PM, Domenic Denicola wrote:
>
>         From: Tim Berners-Lee [mailto:timbl@w3.org <mailto:timbl@w3.org>]
>
>             Since when?
>
>             Is there anything like a public implementation report which
>             tracks that?
>
>
>         https://developer.mozilla.org/__en-US/docs/Web/API/URL#__Browser_compatibility
>         <https://developer.mozilla.org/en-US/docs/Web/API/URL#Browser_compatibility>
>
>         (Chrome 32, Firefox 26)
>
>
>     That page describes URLUtils.pathname as "a DOMString containing an
>     initial '/' followed by the path of the URL."
>
> The high order bit on the web is "is this feature implemented
> interoperably in multiple browsers (and specifically, the browsers I
> care about supporting)". Unlike standards like USB, which are designed
> and implemented atomically, the web platform is implemented a bit at a
> time, by a browser at a time.

It isn't a waterfall process.  Every browser I tested implements 
pathname, if by implements all you mean is that it exposes a property by 
that name.

The problem is that given the same set of inputs, browsers don't 
consistently produce the same values as each other for the property.

This isn't a simple matter of "somebody writes it down, and then 
everybody catches up".  There needs to be a feedback loop involved.

Meanwhile, everybody -- implementers and content producers alike -- 
benefit from a more accurate statements which distinguish between likely 
to be stable (by virtual of compatibility with existing content or 
demonstrable interoperability) from those parts which are experimental 
and not widely reviewed.

> For users of the platform, "I tested it and it runs unprefixed on the
> browsers I care about" is a *far* better measure of stability than "it
> is marked as HTML5" or "it has reached CR".

You seem to be responding to points that I am not making.

> In theory, features should advance along the standards process at the
> same time as they reach interoperable stability. In practice, changes in
> the platform are extremely granular, and the process of advancing
> features along the process involves many details that do not map onto
> practical stability.
>
> I am strongly in favor of better stability signaling in the WHATWG
> process, but at least it gets the basics of stability on the web right.

Here you are drawing a conclusion that is not consistent with the data 
that I have provided.

>     Here are test cases where Firefox 32 and Chrome 37 produce different
>     results, neither of which start with an initial '/':
>
>     http://intertwingly.net/__stories/2014/10/05/urltest-__results/8e738a5350
>     <http://intertwingly.net/stories/2014/10/05/urltest-results/8e738a5350>
>     http://intertwingly.net/__stories/2014/10/05/urltest-__results/010d3b8e54
>     <http://intertwingly.net/stories/2014/10/05/urltest-results/010d3b8e54>
>     http://intertwingly.net/__stories/2014/10/05/urltest-__results/f95c57aa35
>     <http://intertwingly.net/stories/2014/10/05/urltest-results/f95c57aa35>
>
>     There are also plenty of examples where the URL is badly formed and
>     an empty string is returned by both.  A few examples:
>
>     http://intertwingly.net/__stories/2014/10/05/urltest-__results/8e4dba714b
>     <http://intertwingly.net/stories/2014/10/05/urltest-results/8e4dba714b>
>     http://intertwingly.net/__stories/2014/10/05/urltest-__results/a7e821cc81
>     <http://intertwingly.net/stories/2014/10/05/urltest-results/a7e821cc81>
>     http://intertwingly.net/__stories/2014/10/05/urltest-__results/aa0a198c57
>     <http://intertwingly.net/stories/2014/10/05/urltest-results/aa0a198c57>
>     http://intertwingly.net/__stories/2014/10/05/urltest-__results/723aa80622
>     <http://intertwingly.net/stories/2014/10/05/urltest-results/723aa80622>
>
>             What about other browsers?  Do they have plans?
>
>
>         Yes, IE has it marked as "Under Consideration."
>
>
>     The developer.mozilla.org <http://developer.mozilla.org> page
>     indicates that Chrome, Firefox, and IE has "basic support", without
>     defining what that means.  I will note that out of the 256 tests
>     defined for URL spec, there isn't a single one where those three
>     browsers return the same value for pathname.
>
>     Before anybody attempts to infer what point I am trying to make,
>     I'll make it clear:
>
>     * Readers of this page will be done a disservice in that the
>     information isn't technically accurate nor does it adequately
>     capture the state of interop.
>
>     - Sam Ruby
>
>     P.S.  I didn't pick this as an example.
>
>     P.P.S.  For those who want to play with this data, I've placed the
>     current results in JSON for at:
>
>     http://intertwingly.net/__stories/2014/10/09/urltest-__results.json
>     <http://intertwingly.net/stories/2014/10/09/urltest-results.json>

- Sam Ruby

Received on Thursday, 9 October 2014 21:13:44 UTC