Re: Draft [URL] reference update to informative text

On 10/09/2014 11:46 AM, Anne van Kesteren wrote:
> On Thu, Oct 9, 2014 at 5:37 PM, Sam Ruby <rubys@intertwingly.net> wrote:
>> Based on these results, I believe that it is very premature to tell
>> implementors that the spec is ready to be implemented, as they may very well
>> find themselves implementing something that has not yet had wide review, and
>> may very well change.
>
> To be clear, none of what you found is surprising. I ran the same
> tests while writing the specification. (Except for some of the
> recently added ones which appear to be buggy as I pointed out
> elsewhere.)
>
> There's clearly some variation among user agents for these edge cases,
> but at the end of the day they have to converge and that requires an
> attempt to implement the URL Standard.

I agree that convergence is desirable.  I don't believe that "requires 
an attempt to implement the URL Standard." follows from that, however.

As a concrete example: as near as I can tell, the following hasn't had 
wide review, and therefore I suggest that implementers would want to 
consider carefully before they chose to implement it:

http://intertwingly.net/stories/2014/10/05/urltest-results/4b60e32190

I recommend that this be considered an open issue, but you initially 
pushed back, then later appeared to agree:

http://intertwingly.net/blog/2014/09/16/The-URL-Mess#c1412273875
http://intertwingly.net/blog/2014/09/16/The-URL-Mess#c1412338531

There was further discussion on the whatwg lists, but apparently the 
archives are down.  Meanwhile, I'll paste below what you said:

> I also found out today that
> https://github.com/w3c/web-platform-tests/commit/2712b5611a4e048e04a7dc814a7a31413d2d367a
> has not been checked against the specification. So all those tests are
> likely wrong. File URLs differ per browser and per platform so there
> will always be differences there. I tried to come up with a sensible
> algorithm, but it might need checking again.
>
> Overall, the URL specification could likely use another month of
> serious work on the parser side. It is not clear to me now would be
> good to invest that time. I would like to see some more interest from
> implementers first.

I also don't think "all those tests are likely wrong" is a fair conclusion.

- Sam Ruby

Received on Thursday, 9 October 2014 16:06:35 UTC