- From: Harry Halpin <hhalpin@w3.org>
- Date: Mon, 28 Sep 2015 17:42:33 -0400
- To: "public-webcrypto@w3.org" <public-webcrypto@w3.org>
Today, I discussed this issue with Ralph Swick of W3C. It was also
briefly ran by Wendy.
In essence, the problem is what we do we mean by "interoperable
implementation." According to the W3C, the origin of the test
requirement is two-fold: To assure developers what features they can
depend on, and also to make sure two different implementers can read the
same spec and produce the implementations that would pass the same
test-suite. Since the 'browser profile' deals with the former, let's
concentrate on what 'different implementers' means.
We believe it's safe to say that two *different* browser vendors count
as two different implementers. So, if the same algorithm is implemented
twice on Firefox, i.e. on Firefox for Mac and Firefox for Linux, it
would not count. However, if a single algorithm was implemented by
Chrome on Windows *and* Firefox on Windows, since they are different two
different teams, that would count as two different implementations.
This should allow more algorithms to be covered in a liberal sense than
my earlier take on things, but not at the cost of covering algorithms
that have only a single browser implementing. The browser profile will
remain conservative and demand coverage on *every eligible browser*
across *all platforms.*
I'll take an action to update the table to take this into consideration
ASAP.
cheers,
harry
Received on Monday, 28 September 2015 21:42:35 UTC