Re: Proposal to introduce test suite curators

On Fri, Mar 3, 2017 at 10:24 PM, Gérard Talbot <www-style@gtalbot.org> wrote:
> Le 2017-03-03 01:33, Koji Ishii a écrit :
>> Years later, when someone try to implement, find questions to tests, or
>> tests don't look correct, but test authors tend to forget what it was, do
>> not respond, or mail returned in error.
>
>
> Lack of response - indeed - is a problem, has become a bigger problem and is
> one consequence of lack of review within a normal timeframe.
>
> Somewhere someone (I can't remember when, who and in which mailing list)
> said that there is no reason why a reviewer should not make corrections to a
> (presumed incorrect) test that he did not author. I can not reconcile myself
> with such policy. Many years ago, in a documentation project, I remember
> making some harmless but reasonable changes in a file and this made the
> author angry, possibly very angry, at me. It is always preferable, best or
> better to seek and get the agreement, approval of the test author in the
> first place; in the process, this will help the test author to become better
> at authoring tests ... otherwise it will make the reviewer better at
> reviewing tests. Nobody is infallible.

That will have been me.

Ultimately, situations like those that you outline happen as a result
of culture within a project: obviously we cannot dictate culture, but
given what is true in web-platform-tests and elsewhere that browser
vendors heavily contribute to (and, ultimately, the test suites are
largely developed by the vendors), it seems highly unlikely that the
culture around web-platform-tests will change as a result of the
merge.

Some of this is solved by everything being review-then-commit:
assuming people stay on top of the incoming review pile (and hopefully
this will get easier in the coming months), we shouldn't have absentee
authors to outstanding, new, tests. (At the same time, we saw this
with TTWF: the day after the event, nobody would fix any bugs in their
tests any more.)

Similarly, someone reviewing an existing test in the repo will still
be pushing to somewhere where their fixes will be reviewed before
being merged, and as such that should stop any feeling of toes being
trampled on: someone, after all, has the review and approve the fix
the reviewer made.

Equally relevant is the fact that test authors move on: if you take a
random test out of the 2.1 testsuite, there's probably a reasonable
chance the author no longer has much to do with CSS on a day-to-day
basis. Given the large number of those that never got properly
reviewed, trying to do review now and contacting the author now is
likely an effort in futility: many of the given email addresses will
simply no longer exist as a result of the person changing employer.

/g

Received on Tuesday, 14 March 2017 17:19:24 UTC