Should OARS be concerned about meta-data in content created by applications?

[schumaml]

Coming here from https://gitlab.gnome.org/GNOME/gimp/merge_requests/25

When going through the questionnaire to create a suitable rating, I noticed that there are some questions concerned with sharing personal data.

With most application which create content, this can also happen by adding meta data to the files involved. Should this be covered by OARS as well?

As an example, sharing images with your past GPS locations embedded can be as good as real-time location data if it enables others to predict your location at a specific time.

[hughsie]

This is a tricky question. I don't think that files created by the application can be covered realistically, as GIMP can be used for anything from create a school logo or face-swap porn images.

Adding metadata is something that maybe does require some thinking about. If the app is adding rather than just processing some existing metadata I'd argue that's adding personal information and is covered by the existing section. Does that help? I'm happy to make changes to the generator if required.

[schumaml]

Well, it mostly depends on how OARS will be promoted, I guess... as an example, what expectations do you want to e.g. a parent have about an application which has an empty OARS rating element?

Am I right to assume that metadata has not been part of the concept yet at all?

[hughsie]

what expectations do you want to e.g. a parent have about an application which has an empty OARS rating element?

Then it's suitable for all ages. It's a bit like a pen and piece of paper -- you can use it to draw scribbles or you can write out some crazy satanist ideology -- but you wouldn't put an "18" rating on a pen.

metadata has not been part of the concept yet at all

Not really. I agree that handling metadata (e.g. where the photo was taken) should be considered, but I don't know if it's in scope for OARS.

[schumaml]

It's not about the pen - the paper has your address written on it in invisible ink, and it is about whether you expect tools to highlight this and remove it before it gets sent to the public.

On the more active side, a metadata dialog alone could encourage some users to fill in the data according to the labels. and this would effectively be like answering an asl? query in an online chat.

And when we get to images, there have been some cases where embedded thumbnails did show quite a lot more than the cropped portrait, and this was because the software involved did mot modify the metadata.

[hughsie]

So it's less "app includes location data" and more "app doesn't strip existing location data".

[Jehan]

So apart from discussions on updating the spec, what is the current status of OARS? Someone made us a patch which basically just add an empty <content_rating type="oars-1.1" /> in our appdata.

Does it make sense? Should we just accept it? I seriously have no idea what to make with it.

@schumaml actually raises quite serious questions about the power (and risks) of metadata. But of course, it is mostly stuff from data we import, as you note (though it is also possible to create metadata explicitly from within GIMP), and I am also unsure if this is the same scope of issues.

@hughsie Anyway personally I am mostly looking for your proposition on how to deal with this patch and whether or not we should accept it. Thanks!

[Jehan]

@hughsie Rewording a bit my previous message: should we just set our OARS this way (as submitted as a patch by someone on our bug tracker):

  <content_rating type="oars-1.1" />

Or this way:

  <content_rating type="oars-1.1">
    <content_attribute id="social-info">intense</content_attribute>
    <content_attribute id="social-location">intense</content_attribute>
    <content_attribute id="social-contacts">intense</content_attribute>
  </content_rating>

As proposed by schumaml, considering indeed that images created by GIMP may indeed leak personal information (full name, company…), and even location (GPS or address), or contact (email, full address, URL, phone numbers, etc.), and more. These leaked info can either be just passed along (not deleted) from other software and they can also be created from within GIMP itself (you can edit/create metadata since GIMP 2.10.0).

Also we all know that people don't really fully understand when they fill out their info in software.

So is OARS current scope (at least in v 1.1, not talking about possible future updates of the spec) to also care about such metadata concerns potentially leaked by media created by the program?

[Jehan]

P.S.: merry Christmas!

[hughsie]

OARS was primarily designed for computer games, and using it outside of this for all applications is going to have a slight impedance mismatch. The OARS rating is used to assign a suitable age category, so using social-*=intense is going to give GIMP an 18 rating in GNOME Software. This is clearly wrong. I think the right thing to use is <content_rating type="oars-1.1" />

[Jehan]

OARS was primarily designed for computer games, and using it outside of this for all applications is going to have a slight impedance mismatch.

Well this was actually my very first interrogation. Is that OARS rating even meaning something in GIMP. If you think this is just off-topic for us, we could just refuse the patch. We want to do the right thing if possible. :P

So do you think it is even more meaningful to just not add a <content_rating> tag in GIMP? Or we should add the empty one? (last question, seriously! Then I either merge or reject the patch according to your last answer)

[hughsie]

Or we should add the empty one?

An empty OARS rating would be my choice. Thanks.

[Jehan]

Ok I merged on our side.

I would suggest adding some concept of metadata (and its risk) for a OARS 1.2 or something. So maybe this report could be used for this. Though I don't know, maybe you consider this completely off-topic with OARS spec.

I let you decide and keep or close this report. :-) Thanks for all and happy new year!