Help needed for setting permissions correctly so observations can be uploaded to Atlas of Living Australia

[environ]

Yeah, that’s the only page I’ve found each time I’ve looked - and they do put up an SEP field around NC, but don’t say what they don’t like about SA …

I agree ND isn’t very suitable for this sort of use - it’s more appropriate for “Finished” works of art - though even then I think it says more about the copyright holder than the Value of the work - so aside from perhaps ‘protecting’ some extremely artistic image that someone is precious about, it effectively means you’re saying I can’t even comment on or add an ID to your observation, because that would create a derivative work. Not terribly useful for a community refined data set.

[ I’m similarly a bit surprised inat even allows people to select “all rights reserved” - because that means even inat doesn’t have any right to redistribute or let users make additions, except if you argue that the inat ToS somehow has precedence over the user’s explicitly chosen licence …]

But SA doesn’t limit what you can aggregate with it or what licences those things can be under - it even doesn’t prohibit you from saying “my original contributions have licence X” in a derived work, as long as people who receive it are also free to redistribute it under the same terms as the original creator did. That’s why it’s ok for inat to include work under all the licences it does, and why it’s ok for ALA to import all of that work (and it rides the fence of being the Australian node for both inat and GBIF :) All of those datasets include machine readable licences now, so any user that might be a problem for can filter for licences they like themselves anyway.

That’s why I’m curious what GBIF’s actual rationale is … all I can think of that might sort of make sense is that the lawyers for the kind of publishers who traditionally required researchers to assign their copyright to the publisher before their paper would be included in an expensive subscription-only journal thought they might be able to argue that was “non-commerical” because they never pay the people who did the work to produce the paper - but worried that SA would give people (including the original authors) a clear right to continue to publicly share the paper after that assignment …

I don’t know enough about GBIF though to know why they might care about that, so I kind of hoped someone who does might see this thread :)

Without a good reason to think otherwise I’m still inclined to think and tell people that if GBIF doesn’t like SA licenced work from “community” contributors, then that’s their problem not ours …

[spiphany]

@environ

But SA doesn’t limit what you can aggregate with it or what licences those things can be under - it even doesn’t prohibit you from saying “my original contributions have licence X” in a derived work, as long as people who receive it are also free to redistribute it under the same terms as the original creator did.

But if you are using many observations which were made available under a variety of licenses, it becomes difficult to “distribute under the same terms” if the terms are not the same for all the data. It may not be feasible to individually list each observation and the license, or to determine what license should be applied to your derivative work (for example, your analysis of the dataset).

@environ

[ND] effectively means you’re saying I can’t even comment on or add an ID to your observation, because that would create a derivative work.

iNat actually has two separate licenses for observations: one for the media (photos/audio) and one for the observation data, although this is not displayed in a way that is particularly obvious.

I don’t believe that comments/IDs/annotations added to the observation would be considered modification if the “observation data” is understood as the facts of the observation (place/time/observer). Comments by other users are clearly not the intellectual property of the observer; I suppose the ID (what was seen) is more debatable.

Note that there are important unresolved questions about whether observation data can even be copyrighted in the first place.

[environ]

@spiphany

if you are using many observations which were made available under a variety of licenses, it becomes difficult to “distribute under the same terms” if the terms are not the same for all the data.

That’s only a problem if you actually combine differently licenced things to produce your derivative work - otherwise you are still free to distribute them as a multiply-licenced aggregation. And realistically, as long as we live in a culture that believes in both property rights and online over-sharing, that’s a problem that doesn’t go away just because any single source declares only a subset of exercisable property rights to be ‘acceptable’.

Every user of third party data, for anything other than their own private use, needs to do their due diligence on the licence terms of everything they use, right down to the fonts they typeset their paper with. The GBIF ToU even tell you as much due to their data containing a mix including NC licenced data. And yes, it can be a horrible minefield given we only have a loose approximation of globally applicable rules for what is or isn’t acceptable - but it is the world most of us currently live in.

@spiphany

It may not be feasible to individually list each observation and the license, or to determine what license should be applied to your derivative work (for example, your analysis of the dataset).

Including machine readable licence statements in the data (as inat and ALA etc. do) should make it trivial to filter for only data that is acceptably licenced for your use - or at least more so than filtering the other fields for acceptable data quality.

And I wouldn’t say that analysis of the data automatically, or always, makes your analysis a derivative work. In the world of software, where copyright and licencing strongly apply - a legally accepted way to resolve licencing incompatibility is via Clean Room Design - where one team analyses the incompatible component and writes a specification, which is then unencumbered by the original licencing - and then gives that specification to another team to reimplement without any other knowledge of the original protected property which would make their work a derivative.

I can write a review of a protected movie I saw in the theatre without it being a derivative work needing to be compatibly licenced with the move itself. But if I take a clip of that movie and publish it with my own voice overs, that almost certainly is.

This is why ND on images is troublesome - it means I couldn’t include them with my paper if I drew an arrow on top to point to some thing of interest, or cropped an inset image to show close detail of that portion, or possibly even if I captioned it etc.

I agree the issue of the actual observation data is an even more tangled mess of things like “facts aren’t copyrightable” and “but the way they are expressed may be”. And notes I make on my own obs may or may not be sufficiently original and creative …

For genuine research, if you live in the right legal circles, the doctrine of “fair use” also protects your analysis, at least so long as the character of your analysis is analysis and not simple plagiarism or a mash up of other people’s work.

@spiphany

iNat actually has two separate licenses for observations

fwiw, I think it’s actually three - iirc you can licence audio differently to images, though that possibly just makes the accidentally awkward permutations even more insidious :)

Anyhow, I’m not looking to contest anything you’ve said, and do appreciate your thoughts on this - you clearly understand this better than most people care to and we surely could licence-wonk the terrible little details of this until everyone else’s eyes glaze over - this is just why I (still) don’t understand GBIF’s decision to pre-filter data along relatively arbitrary lines, when it doesn’t actually remove the problem for end users, it just removes their ability to decide for themselves which subset of (for some uses) mutually incompatible data is the portion that effects their analysis the least to remove.

I’m clearly still missing something about what they were thinking might be a problem for them that isn’t a problem for other data collections, or in other similar circles aggregating things where licence choice is even less constrained than it is on inat. It’s a minor consolation that it is now looking like it’s not something that should have been blindingly obvious :D