Re: test suite meta data

On 2013-07-31 19:32, Rebecca Hauck wrote:
> On 7/31/13 10:30 AM, "James Graham" <james@hoppipolla.co.uk> wrote:
> 
>> On 2013-07-31 14:12, Arthur Barstow wrote:
>>> On 7/31/13 8:54 AM, ext Tobie Langel wrote:
>>>> Hi folks,
>>>> 
>>>> Turns out I need info like GitHub handle of the test coordinator 
>>>> for
>>>> a spec, mailing list to send pull requests to for each spec, etc.
>>>> 
>>>> I can either put all of this information along with the scripts I'm
>>>> writing, or we can put it directly in the repository in each
>>>> sub-directory.
>>>> 
>>>> So for example, you'd have a manifest.json file in FileAPI/ which
>>>> would contain the following info:
>>>> 
>>>> {
>>>>      "test-coordinator": "KrisKrueger",
>>>>      "mailing-list": [
>>>>          "public-webapps-testsuite@w3.org"
>>>>      ]
>>>> }
>>>> 
>>>> Thoughts?
>>> 
>>> I tend to prefer what you describe above i.e. to distribute this 
>>> info
>>> in each toplevel directory (e.g. FileAPI/).
>> 
>> That seems fine, but I wonder what you plan to use this for. I can't
>> imagine all the subscribers to public-webapps-testsuite will want to 
>> get
>> review spam for every webapps review, with no real way to silence it 
>> for
>> reviews they aren't following.
> 
> I suspect Tobie is in the process of implementing the proposal he 
> floated
> to us little while ago [1]. You're right that you can't silence the 
> mails,
> but with subject prefix he proposed (e.g. [shortname]), they're easily
> filtered. I'd suggest the the prefix includes a keyword that clearly
> indicates it's PR activity (e.g. [github-shortname] or
> [github][shortname]).  Would that be acceptable?

My point is that a specific review can often generate dozens of mails 
(in no small part due to the way that github insists on one mail per 
comment). Getting all these mails for a review that you are not 
interested in is tedious, even if you are interested in other reviews 
for the same spec. I don't see how it is possible to opt-out of these 
mails on a per-review basis in the proposed system. Based on experience 
of review systems in general - even less spammy ones than github - 
people will quickly become annoyed by this.

It is possible that I misunderstood and that tobie is proposing one 
email per review to the mailing list, which is OK, but does make it hard 
to follow activity on reviews you do care about (I don't know, but I 
guess it must be possible to subscribe on a per-review basis using 
github alone?).

> On a somewhat related note, right now I happen to be editing the doc 
> that
> describes how to notify people of new test submissions [2]. It 
> suggests
> (optionally) notifying public-test-infra or #testing in IRC (as listed 
> in
> the Communication Channels doc [3]). Should this be changed?

Well I at least currently get multiple emails per PR, so one more to 
public-test-infra isn't that helpful. Asking questions on #testing is 
fine of course.

Received on Wednesday, 31 July 2013 19:29:49 UTC