License field

[m-mohr]

I am just comparing standards and how they handle certain aspects, e.g. the license field. When looking into STAC I found that there are several inconsistencies in the documentation.

In the JSON spec it says:

Item's license name based on SPDX License List or following guidelines for non-SPDX licenses

extensions/transaction/transaction-fragment.yaml / json-spec/json-schema/stac-item.json / api-spec/STAC-fragment.yaml define the files as follows:

Data license name based on SPDX License List

The JSON spec does not mention what the guidelines for non-SPDX licenses are and looking through the SPDX web page I couldn't really find an answer. Additionally, the other mentioned schemas in the repository don't mention anything other than SPDX licenses. So what to do for proprietary licenses?

The catalog (static-catalog/json-schema/catalog.json) even specifies a new JSON object for licenses, which can be used instead of the the SPDX licenses. It doesn't look to be well thought out though, e.g. shortname of "format": "email" sounds confusing:

    "license": {
      "type": "object",
      "properties": {
        "name": {
          "type": "string"
        },
        "shortname": {
          "type": "string",
          "format": "email"
        },
        "copyright": {
          "type": "string"
        },
        "link": {
          "type": "string",
          "format": "uri"
        }
      },
      "required": ["name", "link"]
    },

Another question regarding the license in the catalog: Is it the data license or the catalog metadata license?

One of the examples uses it completely different: json-spec/examples/digitalglobe-sample.json contains (C) COPYRIGHT 2016 DigitalGlobe, Inc., Longmont CO USA 80503.

That should be sorted out and be made uniform across the spec.

Suggestion / Alternative 1: I'd suggest to change the license definition and make it an object.

Fields:

• name (string): Full name of the license.
• spdx (string): If available, license identifier from the SPDX License List. (Adopted from how epsg codes are implemented)
• url (string): A URL to the license text. MUST be in the format of a URL.

Example for SPDX licenses:

{
  "spdx": "Apache-2.0"
}

Or a longer form, replicating information from the SPDX list:

{
  "name": "Apache License 2.0",
  "spdx": "Apache-2.0",
  "url": "http://www.apache.org/licenses/LICENSE-2.0"
}

Example for non-SPDX licenses:

{
  "name": "CeCILL-B Free Software License",
  "url": "http://www.cecill.info/licences/Licence_CeCILL-B_V1-en.html"
}

One thing that bothers me here is that the url (or should it be link?) could also be in the links section with the rel type "license", but that would spread information about licensing in two different places.

Alternative 2: If we don't want to change the current specification much, but want to be flexible: Allow only SPDX license identifiers in the license field and require users to add a link with rel license to the links for other licenses.

What do other specifications/standards do?

• OpenAPI has a JSON object containing a required free-text name and a optional url to the license text.
• DCAT / Dublin Core: They have license and rights. license holds a reference to "a legal document giving official permission to do something with a Resource". rights seems to be free text. Note from the docs:

  dct:license, which is a sub-property of dct:rights, can be used to link a distribution to a license document. However, dct:rights allows linking to a rights statement that can include licensing information as well as other information that supplements the licence such as attribution.

• ISO 19115: They want to kill it with a whole class MD_LegalConstraints, which holds several properties, including access and usage constraints, which itself are classes of type MD_RestrictionCode, which have actually 8 properties: copyright, patent, patentPending, trademark, license, intellectualPropertyRights, restricted, otherRestrictions. Seems to be completely over the top. Mostly free text and seems to result in overly long "expressions" (Example).
• OGC CSW 3.0: Named rights and doesn't really cover the license. It is simply a string with "Information about rights held in and over the resource". Example from the spec: "Copyright © 2011, State of Texas"
• OGS OpenSearch EO: uses ISO 19115
• INSPIRE: free text, uses to ISO 19115 and ISO 19139, recommends to include a link.

[m-mohr]

Another interesting idea: What about allowing an array of licenses to be specified? I just saw the Planet Disaster Data Catalog and it includes:

Imagery is provided under Creative Commons licenses, free of charge, with either CC-BY-SA or CC-BY-NC.

For dual-licensed data, there is no good way to describe them in STAC, but it could be an array:

{
  "name": "Planet Disaster Data",
  "license": [
    "CC-BY-SA",
    "CC-BY-NC"
  ],
  ...
}

[mojodna]

SPDX supports multiple licenses via AND and/or OR, so perhaps that's sufficient?

[m-mohr]

This was discussed in STAC spring #3.

We concluded to allow SPDX licenses in the license (or license_name?) field and add an additional license_url field. license_url should be set only when license is set to "proprietary". I think license should just be a valid SPDX expression to allow multiple licenses. I am not sure how this incorparates with our custom term "proprietary" though. SPDX uses the strange name noassertion for that behaviour.

Some interesting SPDX tools I just found:

• PHP library: https://github.com/composer/spdx-licenses
• Java library: https://github.com/tbouron/spdx-license-checker
• JS license list: https://www.npmjs.com/package/spdx-license-list / https://www.npmjs.com/package/spdx-license-ids
• JS SPDX expression parser: https://www.npmjs.com/package/spdx-expression-parse
• Python tools: https://pypi.org/project/spdx-tools/

These tools are probably mostly interesting for clients, e.g. STAC browser.

Others handling that issue:

• proprietary vs. SPDX native wording (noassertion / none): https://github.com/composer/composer/issues/2152
• npm uses UNLICENSED instead of proprietary: https://github.com/npm/npm/issues/8918

PR still to be done by @jeffnaus or me.

[matthewhanson]

So now I'm wondering if we need license_url at all, and instead can just include license as a link with rel="license"

[cholmes]

I think I like link with rel=license. Especially if we do have a license name as a property.

And yes, I think we start with 'proprietary' for non-SPDX license. I don't think most imagery is 'unlicensed', it's in non-free license. We might choose a subset of SPDX - just the Creative Commons, odbl and public access ones, since software licenses aren't really applicable.

We could also say 'custom' and require that the link to the license is there if you use 'custom'. Eventually may be good to say some 'types' of proprietary licenses, like 'derivatives allowed'. But defining that list is far out of the scope of STAC (I've wanted to work on it before, since it is really needed)

[m-mohr]

I do see advantages and disadvantages in the rel=license thing:

• Advantages: consistency, no special field for license urls, can easily be followed by clients.
• Disadvantages: Two related information are placed at different locations, creating a simple link in tooling (e.g. stac browser) is more difficult. You need to find the rel=license first in the list of links.

+1 on proprietary

+1 on SPDX licenses, but definitely -1 on subset of SPDX licenses within STAC. This complicates things and we need special treatment in the spec, which should be avoided to keep things simple. Licenses not applicable will just not be used anyway. Most times the license is set before making a STAC item so users will not go through the SPDX and choose one, but just look for their pre-defined license identifier. I don't see any benefit on sub-setting.

[matthewhanson]

I think there will be other cases where info is split up so I don't see that as much of a problem.

With Datasets we'd probably want a Dataset field indicating the dataset it belongs to, but then also a link to it, so it's a similar case.

[m-mohr]

Either I don't understand your last paragraph or there are some "dataset" that should be replaced by "license"... I guess you are basically saying: Make Item and Dataset consistent regarding how to handle licenses and their URLs.

[cholmes]

Ah, good point on SPDX licenses. I wasn't thinking of doing a full constraint on them. Maybe just more a 'recommendation' - these are good ones that are relevant to geospatial information.

As for the disadvantage, @mojodna seemed less compelled by arguments that 'it's harder for tooling' (like links vs dicts), since it is just a bit of code that has to be written once.

[m-mohr]

Sure, it is not a big thing. I'll make a PR next week.

[m-mohr]

Is included in the restructuring PR #202.

[m-mohr]

This should be completed with the latest PR.