Single unified License for works derived from Smoke Detector Data.

[magisch]

This came up starting here: http://chat.stackexchange.com/transcript/message/37245230#37245230 And then here again: http://chat.stackexchange.com/transcript/message/37248760#37248760

Basically my proposal is to make a commitment similar to what we did with repository contributions that all our works utilizing smoke detector's collected data (like @AWegnerGitHub 's fancy graphs or analytics models we make based on that data) be licensed under a single license similar to Smokey itself, in order to facilitate and enable easy reuse of those works by others.

One of the questions regarding this would be which license to use, and another of what scope of work it'll cover.

[Undo1]

So you're proposing a 'cancer license', like GNU? Realistically, I tend to oppose things that force folks to go public and allow modification of what they've done. For example, I have work with Smokey data that's been merged with data from moderators and staff. I can't make that public.

Same with the MS database - I can't just copyleft the whole thing.

In the end, it's the normal question: Is this a real issue?

If it is, I'd probably recommend going with a CC0-type license. If there's one out there that allows use, but doesn't force a license upon others, I'd probably be happy. Maybe a CC0 with a "derivatives default to a permissive license, but you can specify otherwise if you want".

[ArtOfCode-]

Limitations:

• We can't force a license on past work that isn't our own - it's just not possible. Andy, for example, would legally still have to agree to a copyleft license if we wanted his work to be available under a public license.
• Copyleft just isn't friendly. I'd strongly recommend staying well away and just going with something permissive.

@angussidney pointed out the ODC PDDL or BY licenses. I'd be happy with the BY license - it's a license designed for datasets/databases, and it requires some form of attribution without being copyleft like the GPL or BY-SA.

[AWegnerGitHub]

I need a bit more information before forming an opinion.

• What is the end goal of this?
• What would be required to be released? As I understand it, post content is already CC by SA because it comes from SE. RegExes and Blacklist information is already Apache/MIT because it's code (until that is moved to the database, then it's not code any more...). What could this cover?
• I play with MS data frequently. I post pretty pictures to Stack Exchange's chat rooms if I find something interesting. At this point, it's CC by SA, correct? What does adding a license here do or require at that point? Do I also have to release the actual raw data I used to generate the pretty pictures, or is my text explanation enough (which is also on SE).
• What is required of the data I play with but don't post to SE? Does a license require that I now release that?

[ArtOfCode-]

@AWegnerGitHub Responses generally depend on which license we choose; for the purposes of this I'm going to assume ODC BY.

• End goal: Might help to read the original chat conversations on this (linked in the issue) to clear that up a bit; I think it's just to make our data more permissive/available to other people to use. Our software is already available, but our data is only available in limited form through metasmoke or data dumps (where the licensing status is ambiguous, so using them for any major project without confirming with us first would be a bad idea).
• Things to release: complex. Post content is BY-SA. Anything in our repos is dual MIT/Apache. Beyond that, we can pretty much release what we like under ODC BY, though if we were doing selective releases we'd need a release dump process.
• Whatever you post to SE is licensed to SE under BY-SA (or CC-wiki, not sure if that still applies in chat). So, if you post a PNG of a graph you've generated in chat, that image is licensed under BY-SA. The data you used to generate the graph is not. If we licensed under ODC BY, technically you'd have to add an attribution note ("graph generated using data from Charcoal"), but I doubt any of us would hold you to that.
• Nope. Mainly, since you're doing it yourself without really telling anyone, licenses are totally unenforceable - but also that's not something the license requires anyway.

[magisch]

@AWegnerGitHub The end goal would be to have every derivative work we release (so you choose what you post) on GH or somewhere else available under a single license so people can reuse it.

[AWegnerGitHub]

@magisch That seems unnecessarily burdensome.

As an example, if I perform some kind of analysis and post it, this now requires that I abide by the terms of SE (because I posted it to the chatroom) and whatever terms we come up with here. Why the need for two separate licenses? I'm already bound by the CC By SA license once it's posted over the SE.

What are you expecting people to "reuse"? An image? Great...CC By SA, because it's already under that license. An explanation of the image? Great...already under CC by SA. Now, we add one of the licenses mentioned by @ArtOfCode- (which I don't have a link to...) and we have added more that someone using the data needs to do:

• I generate stuff and post it to SE. "This data is generated using data from Charcoal, which in term uses data from Stack Exchange."
• Someone sees my post and wants to reuse it: "Data gathered by Andy, who used data from Charcoal, which uses data from Stack Exchange."

---

If we licensed under ODC BY, technically you'd have to add an attribution note ("graph generated using data from Charcoal"), but I doubt any of us would hold you to that.

You better. That's the point of the license. If you don't, it's worthless. I don't like that we are already saying we aren't going to hold people accountable for a license.

Our software is already available, but our data is only available in limited form through metasmoke or data dumps (where the licensing status is ambiguous, so using them for any major project without confirming with us first would be a bad idea).

The API documentation is hosted in the MetaSmoke repository, a repository with a CC0 license in it, which strongly implies the data retreated via that API is licensed the same way. Smokey has two licenses (Apache and MIT). Adding a 4th license into this mix isn't going to make anything less ambiguous. Especially with activities going on in other repositories and possible backports from NG to Smokey. By taking certain things that are current MIT/Apache licensed out of the code (blacklists, regular expressions) and putting those in the database we've only made things more confusing. Are those no longer code? Are they covered by the Apache/MIT scheme or by the new data scheme? Or, are they under one scheme until a certain date and then changes after that date change the licensing?

---

Either I'm not understanding the need for this and potential confusion it's going to cause, or it's adding restrictions that really don't seem to be needed (or both? neither?), but the way it's described above by @magisch and @ArtOfCode- does not make this seem worth while.

---

[angussidney]

Let's just set something straight here: what we're talking about licensing is anything in the MS DB that isn't already licensed under CC BY-SA because it came from SE. For example: feedback, reasons for which a post was caught, any of the graphs, status/scan rate information from Smokey, autoflagging logs, etc.

What we're not giving a license to derivative works - e.g. Andy's graphs that he made up to illustrate autoflagging - that's up to the owner to decide. We're just giving a license to the raw data itself.

Or at least that's what I'm thinking/talking about. Maybe I just had it wrong.

---

Anyway, my thoughts:

• We need to provide a license for this 'other' content, or otherwise it would be legally ambiguous what license would apply if anyone wants to use our data in their project.
• I agree with Undo - we don't want to go with a copyleft license, a permissive license would be better. CC0 works, or even better, the ODC PDDL, which I mentioned earlier. Derivative works shouldn't be forced to make their works public too - especially when they contain privileged information which can't be distributed.
• Do we wnat attribution? For this, I don't think so. But I'll leave that to the rest of you to decide.
• Like Andy said: if we're going to put a license on this, we should enforce it. There's no point going and adding a license which requires attribution, only to make it optional.

[magisch]

@angussidney I was thinking anything published on our official channels that doesn't have a license yet.

[ArtOfCode-]

We need to provide a license for this 'other' content, or otherwise it would be legally ambiguous what license would apply if anyone wants to use our data in their project.

This is the core of this issue. Our data is not licensed - that includes data retrieved via the metasmoke API. There's an implicit right-to-view there, which includes personal use, so nothing we're doing with it right now is an issue, but if someone wanted to come along and use our API to build on our data, there's no license that they can use it under.

The API documentation is hosted in the MetaSmoke repository, a repository with a CC0 license in it, which strongly implies the data retreated [sic] via that API is licensed the same way.

The documentation is under CC0. Probably. The data isn't, because it's not connected to the repository that is under CC0, and the license doesn't extend to it.

Why we should use yet another license - You know how people complain about SE, particularly SO, using CC BY-SA for code because it's not a code license? In a similar vein, using a code or CC license for a dataset/database is equally bad. If we're licensing this stuff, we should be licensing it using a data license.

[AWegnerGitHub]

I was thinking anything published on our official channels that doesn't have a license yet.

This contradicts your previous explanation:

The end goal would be to have every derivative work we release (so you choose what you post) on GH or somewhere else available under a single license so people can reuse it.

(emphasis is mine)

---

The documentation is under CC0. Probably.

This type of statement does not build confidence that adding a 4th license into our solution will resolve anything at all.

In a similar vein, using a code or CC license for a dataset/database is equally bad. If we're licensing this stuff, we should be licensing it using a data license.

Explain why. Not all of us know our licenses inside and out. What does this data license that has been mentioned repeatedly get us over a "bad" CC license?

[magisch]

This contradicts your previous explanation: @AWegnerGitHub Maybe I didn't make myself clear enough. I was saying the same thing I did afterwards, but more convolutedly. I was talking about data and stuff like graphs using that data that we post on our website or github, or elsewhere in an official capacity as charcoal. Basicly what I thought would be nice would be that someone can see something charcoal made and can go "Cool, I can reuse that" without having to piecemeal investigate licensing issues.

One of your graphs would fall under that for instance (pending your approval of course, nobody can license what isn't theirs) if it was posted somewhere like charcoal-se.org

[ArtOfCode-]

Explain why. Not all of us know our licenses inside and out. What does this data license that has been mentioned repeatedly get us over a "bad" CC license?

I haven't actually looked at either license in huge amounts of detail, but... would you use CC BY-NC-ND for code? Would you use the GPL for a scientific paper? Would you use a fork to eat soup?

The documentation is under CC0. Probably.

I said that because GitHub don't make it entirely clear how repository licenses extend to wikis, but I assume they do.

[Undo1]

Closing for now; comments still welcome.