Add browser removal/addition data guidelines

[sideshowbarker]

This change adds data guidelines on removal of existing browsers from BCD, and addition of new browsers to BCD.

Related: #7238

[queengooborg]

(@ddbeck It seems the auto-labeler didn't catch this PR. Perhaps a problem with the glob?)

[ddbeck]

@vinyldarkscratch I opened #7250 to track the labeler problem.

[ddbeck]

Thanks for opening this, @sideshowbarker. One suggestion in a line comment.

(Like the removal PR, this will likely be blocked until I hear back from Chris. I don't actually anticipate this proposal being rejected though.)

[ddbeck]

Had a few more suggestions, now that I've had some distance from this. Thank you!

[sideshowbarker]

Had a few more suggestions, now that I've had some distance from this.

Doublechecked them all and everything looks good, except that I see we’ve got a lint failure. Will investigate right now

[sideshowbarker]

I see we’ve got a lint failure. Will investigate right now

OK, pushed a fix and we’re now all green again for CI/lint

[Elchi3]

Thanks for coming up with these guidelines. Long overdue to specify this :-)

The original browsers in BCD are based on what the MDN wiki had in the old compat tables which were added by MDN contributors over the years. The wiki had sets of different browsers on different pages, but there was a set of browsers that occurred often. So, BCD came to its original list of browsers via the data migration and avoiding to lose information.

I remember BCD added and removed Servo. I think a removal reason was that there aren't real releases and versions numbers.
I think "published release information" might cover this in the addition guideline, ie. for adding a browser we should require that it actually has releases, so that recording compat data makes sense. (there are other experimental projects that asked to be added to BCD and it wasn't always clear to me if/when they actually have releases). We might want to state this explicitly.

Another removal that I remember is Edge mobile. See #3888. It seems to retroactively meet the removal criteria mentioned here.
Before we agreed to remove it, I sent an inquiry to the Microsoft MDN PAB representatives to comment on the removal. I don't know if it would be worth to mention it in the guideline that it is at least worth to try contacting vendors before removal.

Finally, I'd like to mention that BCD's Opera and Opera Android data seems to (almost) meet our removal criteria. Is there another criteria that makes us keep opera? Is it (historical) marketshare? Something else?

[ddbeck]

@Elchi3 Good points on this, thank you for taking a look.

In general, my original proposal was meant to be a guideline for making a holistic decision, weighted heavily toward what's good for BCD consumers and their readers. Perhaps we could rephrase things to say something like, "BCD owners will consider the following…" to show that it's not a complete or absolute set of criteria. This might help with a few of these other issues.

To your more specific points:

for adding a browser we should require that it actually has releases, so that recording compat data makes sense. (there are other experimental projects that asked to be added to BCD and it wasn't always clear to me if/when they actually have releases). We might want to state this explicitly.

Yeah, I was trying to capture something like this with the published release information proposal. I wanted something a bit more substantial than being able to populate a browser data file. I want to know about release process and plans too. I'm open to rewording this, if you've got any ideas.

I don't know if it would be worth to mention it in the guideline that it is at least worth to try contacting vendors before removal.

I don't want to bind us to this. For the removals we have pending, I'm not even sure who I would contact and I don't think there would be any change in the outcome based on that conversation. I'd welcome the return of those browsers, but it seems preferable in a lot of ways to start from scratch. And if a browser has vendor participants here, I would expect (and ask) for their help on matters related to that browser prior to removal anyway.

I'd like to mention that BCD's Opera and Opera Android data seems to (almost) meet our removal criteria. Is there another criteria that makes us keep opera? Is it (historical) marketshare? Something else?

This is a good point. I think the biggest open question there is downstream: do MDN or caniuse readers care? I think if we adopt these guidelines, then we should open an issue to investigate this. The answer might be to keep it (we do update Opera data rather frequently), but it's not a bad idea to take a look.

[Elchi3]

In general, my original proposal was meant to be a guideline for making a holistic decision, weighted heavily toward what's good for BCD consumers and their readers. Perhaps we could rephrase things to say something like, "BCD owners will consider the following…" to show that it's not a complete or absolute set of criteria. This might help with a few of these other issues.

Makes sense to me. After all, a removal or an addition will be discussed in an issue anyways and arguments will be weighted against each other. The guidelines here provide general guidance and hopefully allow us to consider criteria more fairly for the various requests.

I don't want to bind us to this. For the removals we have pending, I'm not even sure who I would contact and I don't think there would be any change in the outcome based on that conversation. I'd welcome the return of those browsers, but it seems preferable in a lot of ways to start from scratch. And if a browser has vendor participants here, I would expect (and ask) for their help on matters related to that browser prior to removal anyway.

Yep, makes sense. I think we don't need to add anything about it then.

This is a good point. I think the biggest open question there is downstream: do MDN or caniuse readers care? I think if we adopt these guidelines, then we should open an issue to investigate this. The answer might be to keep it (we do update Opera data rather frequently), but it's not a bad idea to take a look.

Yeah, opening an issue for discussion is a must have for adding/removing a browser. I think this is obvious, but if anyone cares strongly, it could also be mentioned in the guideline that there must be a discussion in an issue about the case.

[ddbeck]

I think we're ready to go on this. Thank you, @sideshowbarker and @Elchi3 for making this really solid! 🎉