An open discussion list for topics related to the eduGAIN interfederation service.

[Martin Matthiesen <martin.matthiesen AT csc.fi>]

Hello,

I would like to still add some comments below

----- Original Message -----
> From: "Peter Schober" <peter.schober AT univie.ac.at>
> To: edugain-discuss AT geant.net
> Sent: Thursday, 4 December, 2014 11:55:28
> Subject: Re: [eduGAIN-discuss] eduGAIN and non "academic" IdPs

> * Martin Matthiesen <martin.matthiesen AT csc.fi> [2014-12-03 12:40]:
>> If my contract says: Open for "academics in good standing" without
>> registration (explicitly excluding students) and all parties accept
>> that there are slight differences in definition, I see no
>> problem. We don't need to overengineer here.
> 
> It's doubtful RPs will actually be willing to leave that wide open to
> everyone's interpretation, but if you say so and I believe you're able
> to speak on behalf of the community this use case is for, that's fine.

I made the effort to look up the contract and it says the data in question 
can be made available for educational teaching or research purposes (in 
Finnish "koulutus-, opetus tai tutkimuskäyttöön").
This was  interpreted as EPA=faculty. But since EPA=staff will likely conduct 
research, the audience could be widened.
 
>> > What I'd like to see is a sufficiently detailed writeup of the
>> > requirements of the license agreements in your use case, in a way that
[...]

>> I don't have a copy at hand, and if I had, I would not want to go to
>> that level. The agreement allows for "faculty", it might allow even
>> for "employee".
> 
> Does it include "student"?

No.

> If so can we agree here and now that
> "member" (cf. eduPerson defintion) is good enough?
> From any IDP (representing any number of any kind of institution, as
> long as the institution is eligible for federation membership for a
> given registrar, for all member federations of eduGAIN)?
> 
> Examples from Wikipedia (mentioned by Nicole yesterday at the EWTI
> meeting):
> 
[...]
This did not touch the point. We do not need to agree what an academic is. 
eduGAIN has done a great job by defining a controlled vocabulary for 
eduPersonAffiliation, I assume Feide is closest to the suggested use 
(https://www.feide.no/attribute/edupersonaffiliation) IF all IdPs would 
implement EPA like defined by Feide, any country could define Academic the 
way they wanted, with students (="member"), without (="employee") or even 
stricter (="faculty"). eduGAIN does not need to define that.

> 
> So I would suggest the following (eduPerson affiliation values):
> 
> * Require just "member", if you're fine with students being included
> 
> * Require "member AND NOT student" if you're not fine with students
> 
> * If for some reason that's all unacceptable, require "faculty" and
>  find ways how to deal with those not sending it (e.g. by accepting
>  "staff" instead, as the UKfed and eduID.at use that to include
>  faculty).

I suggested mappings earlier on and got a comment like "been there, done 
that, not going back." And I tend to agree.
 
> You still need a way to deal with subjects from IDPs (representing one
> or more institutions) which are not universities (or publicly funded
> research institutions, not that we tag those today or that the funding
> body is part of your "definition" above).  Does that include their
> employees as well (if not all, which ones)? If there's no attribute by
> which RPs can recognize those -- and eduPerson is not seen as suitable
> for those, see the discussion on MACE-Dir I started for this this week
> -- how would you do vetting of individuals from across the globe?

To me it would suffice that commercial entities cannot set 
eduPersonAffiliation with the attributes mentioned above. In my use case 
individuals with a valid need (students that want to study the corpus) can 
always get access upon request (CLARIN RES license).
 
>> Agreements require interpretation too, as most lawyers would be out
>> of work otherwise. But that in turn requires that these attributes
>> are half-way defined and used as defined. And set in the first
>> place.
> 
> Right. You could require "faculty" (which seems closest) but then

I guess it got lost that that is precisely what I am doing right now. This is 
not a theoretical case:

The corpus is described here: http://urn.fi/urn:nbn:fi:lb-2014052718

And it can be accessed here: https://korp.csc.fi/#lang=en

If you login and have "EPA=faculty", you will be able to select HS.fi and 
search in it. Right now.

> you'd lose out on several whole countries where use of that north
> american term is not common (or used at all).

Yes, it so happened that a professor in Gothenburg tried to login during a 
presentation and was not very happy that it did not work, because she had 
EPA=employee, not faculty.
It is just an attribute. This very language I am using here is not my native 
one but it has it's use.
 
> I have no opinion on whether it's worth trying to get more regions of
> the world aligned to north american terminology and what time frame
> (years, in unsigned integers) you'd find acceptable.

It is just an attribute that says one is a professor, teacher lecturer. Take 
the definition in Feide: faculty = "A teacher (or in higher education: member 
of the scientific staff)" as opposed to staff = "An employee without any 
educational responsibility" I assume most universities have professors, 
lecturers and "employees without educational responsibility". The fact that 
Swamid rejects faculty barred the Swedish professor from using my service. 

> As I said before,
> e.g. in .at we've started the process of suggesting to institutions to

By "institutions" you mean IdP-operators, right?

> differentiate between staff (used here like it is in the UK) and
> faculty (in the above Wikipedia sense). Others may not have such plans
> today.

I don't see this as a problem on the level of Universities. They all have 
professors, lecturers, researchers, students. The problem I see is that 
IdP-Operators and/or National Federations cannot agree on a standard to 
express the mentioned roles.
 
>> What we need is an understanding of a proper process that is a bit
>> more binding than the present one.
> 
> eduGAIN (as a service) is a metadata exchange service between R&E
> federations (mostly, there's also the lightweight policy package),
> built in the assumption that changing member federations' policies are
> out of scope.

Yes, and the result is that EPA=faculty does not work. I understand how hard 
it was to get where we are now, but I think the road (=metadata service) 
needs some roadsigns (=commonly agreed attribute semantics) to be used 
effectively.
Otherwise you built a road to nowhere.

> The eduGAIN and REFEDS communities do coordinate and
> exchange experience and standardize things, but neither are in a
> position to mandate institiutions doing things (or what? Expell them
> if they don't assign "faculty" in their IDM?).

That would be one approach, yes. Standardisation without making sure the 
agreed standards are in use, seems quite fruitless to me.

> I don't see how anyone in the world would ever be in a position to
> force an institution to release data (and even affiliation is PII in
> some jurisdictions) to anyone. "You have to or you can't access the
> service", sure, you can do that today.

I am quite sure no-one outside Finland is shaking in fear about not beeing 
able to offer HS.fi on korp.csc.fi to their local audiences. I don't see 
threatening NRENS to be kicked out of eduGAIN working when most of them are 
not really in it. Finland is no model child either: 4 IdPs and 5 SPs. For 
Finland, I know this will change. The more eduGAIN is used, the more 
standards can be enforced. Before that eduGAIN needs to be used. I have no 
answer either on how to get out of that cycle. It would be a great start if 
eduGAIN starts to realise that there is a problem.
 
> Not sure how to move forward if the above suggestions on attribute
> requirements for your use case are not acceptabe.

EPA=faculty is just fine. If I could rely on it.

Martin