Re: Browser Profile Proposal: Testing implementations for algorithm coverage

21 сент. 2015 г., в 12:17, Harry Halpin <hhalpin@w3.org> написал(а):

> 
> 
> On 09/21/2015 12:14 PM, Alexey Proskuryakov wrote:
>> 20 сент. 2015 г., в 17:36, Harry Halpin <hhalpin@w3.org> написал(а):
>> 
>>> 
>>> On 09/20/2015 03:37 PM, Alexey Proskuryakov wrote:
>>>> Hello Harry,
>>>> 
>>>>> 14 сент. 2015 г., в 12:43, Harry Halpin <hhalpin@w3.org> написал(а):
>>>>> 
>>>>> Apple has been left out
>>>>> until they remove the webkitSubtle prefix and replace it per spec with 
>>>>> 'subtle', which has been communicated to them.
>>>> Could you please clarify why this precludes testing?
>>> We can't have browser developers "checking" to see if they are using
>>> Apple's special prefix. We're happy to test, but we'd prefer you guys to
>>> hurry up and promise to remove the webkitSubtile prefix. Can you do that?
>> WebKit's implementation is known to be based on an older version of the spec which is not compatible with the current one, and I think that it's a good idea to have that detectable from JavaScript. Notably, a developer has a way of knowing that HmacKeyParams.length uses different units - situations like this is exactly what vendor prefixes are for.
>> 
>> WebKit's implementation is real, it ships in real products (and was among the first WebCrypto implementations), so it feels strange to have it excluded based on a technicality. This e-mail thread in particular discusses something that is in no way related to having a prefix - the list of supported algorithms wouldn't change upon removing it.
> 
> Is there any possibility we can simply upgrade WebKit's implementation
> to be compatible with the newer version of the spec?

I agree that this needs to be done, however this certainly can't be done soon enough to affect the data gathered in this thread, which is already ongoing.

> I understand this may not be the highest priority for you, but it means
> a lot to us at W3C to have a consistent spec without vendor prefixes
> before we exit Rec. The entire point of the CR testing, which determines
> the 'browser profile' in terms of algorithms, is to ensure developers
> can use the API across browsers without any browser-specific dependencies.

I think that adding WebKit as it stands now in shipping products to the test matrix would be of good service to developers.

> We know lots of hard work has already been put in, so this final push
> would be great.
> 
>   cheers,
>        harry
> 
>> 
>> - Alexey
>> 
> 
> 

- Alexey

Received on Monday, 21 September 2015 19:29:17 UTC