Re: Test review procedure

On Wed, Mar 16, 2011 at 9:32 PM, Kris Krueger <krisk@microsoft.com> wrote:
> The btoa and atob tests are not part of the spec so they can't be approved.
>
> http://www.w3.org/Bugs/Public/show_bug.cgi?id=12029
>
> Ian 'Hixie' Hickson 2011-02-10 19:44:26 UTC
> Oops, my bad. I didn't mean to include window.atob() in the W3C copy. Will take
> care of that.

See comment 5 there.  They were briefly removed from the draft due to
a misunderstanding, but were quickly re-added, and are part of the
current Editor's Draft:

http://dev.w3.org/html5/spec/webappapis.html#atob

Thus I don't see any standing objection to approving the tests.

> Now for the reflection tests - I'd suggest the following to get consensus.
>
> First refactor the tests so that it tests a few attributes to start rather than the current huge list.
> The way the tests are currently factored it requires every test to be reviewed before any can get approved.

I've already split it up into a number of separate files.  For
instance, the "Metadata elements" file contains only 96 tests, for six
elements:

http://dvcs.w3.org/hg/html/raw-file/tip/tests/submission/AryehGregor/reflection/reflection-metadata.html

I don't expect this will ease review much, though.  reflection.js is
going to be by far the most difficult part to review.  The actual
elements or attributes are just a matter of comparing data files to
the spec, which is simple mechanical work that should take less than
an hour for the whole thing.  I could break up reflection.js, but only
at the cost of making it more complicated, which I don't think would
make review easier overall.

More to the point, I don't notice you or anyone offering to actually
do the work of reviewing the tests in the event I do break them up.
If that's what it takes to get them approved, I'll do it, but I'm not
going to do it just because it might possibly encourage review.

On Thu, Mar 17, 2011 at 1:51 AM, L. David Baron <dbaron@dbaron.org> wrote:
> Why do we still have a review process?  And what does it take to get
> it changed?

I would also like to know how we could go about changing the current
process, procedurally.  It's possible that we won't reach complete
consensus on a change, and in that case we'd need some way to resolve
the dispute.

Received on Thursday, 17 March 2011 18:03:14 UTC