W3C

– DRAFT –
Linked Web Storage

19 May 2025

Attendees

Present
acoburn, AZ, Beau, bendm, cpn, eBremer, ericP, gibsonf1, kaefer3000, pchampin, uvdsl
Regrets
-
Chair
acoburn
Scribe
AZ

Meeting minutes

acoburn: introductions & announcements

<RazaN> +1

Beau: work at technical institutes on social innovations
… background in system integration

Introductions and announcements

Beau: familiar with linked data
… had work on the Solid spec
… worked on the YYY project

acoburn: this experience will be useful for the group

Monsecom: background as a data scientist

<Beau> Will do

RazaN: I'm a postdoc recently joined AZ's group
… will be using Solid in my research
… Semantic Web background

acoburn: we talked about a possible face to face meeting
… we are working on something maybe in the USA in October or November

<gibsonf1> I think its UK or EU (not USA)

<ericP> i'm available

<Monsecom> Beau and I work as developers at VITO (vito.be) , mainly in the We Are (https://we-are-health.be/en) project.

acoburn: there will be a holiday soon in the US and we might skip the meeting

<pchampin> I will be available

<Monsecom> I have a background in computer science

<gibsonf1> available

acoburn: memorial holiday in the US and spring holiday in the UK

<Monsecom> available

acoburn: nobody seems to complain, so we can keep the meeting

<gibsonf1> @AZ on earlier note about meetup, not USA but UK or EU

Action Items

acoburn: Hadrian is away today

Use Cases status summary

acoburn: let us jump to next agendum
… there're 2 PRs open by Pierre Antoine
… I think we should merge that
… this is preventing us from publishing as a working draft

<acoburn> UC title

<gb> Pull Request 155 Add a "real" title to the document (by pchampin)

pchampin: I overlooked the fact that the title was wrong
… which is what hold the publication back

acoburn: seeing no objection I'm going to merge
… another PR open by Hadrian
… it introduces new use cases with serialisation formats
… BTW, this is a introductory list of UCs, no guarantee they will be in the final report
… there are specific UCs
… that were merged or closed
… because they were part of consolidated UCs
… in the published document, we want to have pointers to the GH issues and a trace where they come from
… issue ucs #30 was closed

<gb> Issue 30 not found

acoburn: issue #35 was closed too

<gb> Issue 35 not found

acoburn: reminder: we can find the doc that we want to publish in Github and there is a loink to the final W3C location
… issue #95, #111 were also closed

<gb> Issue 95 not found

<gb> Issue 111 not found

<gb> MERGED Pull Request 123 Updated glossary (by hzbarcea)

acoburn: w3c/lws-ucs#112, w3c/lws-ucs#134 were closed/merged too

<gb> CLOSED Issue 134 [REQ-F] Adding, updating, deleting Resources in Storage (by hzbarcea) [review]

<gb> CLOSED Issue 112 [UC] Storage portability from a provider to another (by hzbarcea) [duplicate] [usecase]

acoburn: questions about these?

<Beau> My GitHub account: https://github.com/beauvangemert1990

Discussion: Authorization

acoburn: we started the conversation about authorization last week
… we can talk high level what we want
… not going into specific solution
… there are several issues related to authorization
… for instance w3c/lws-ucs#70

<gb> Issue 70 [UC] Discoverable Authorization Requirements <DRAFT> (by termontwouter) [triage] [usecase]

acoburn: another is w3c/lws-ucs#67

<gb> Issue 67 [UC] Processing-based Access and Usage Control (by termontwouter) [triage] [usecase]

acoburn: this one is more about what was discussed last week
… another is w3c/lws-ucs#66

<gb> Issue 66 [UC] Purpose-based Access and Usage Control (by termontwouter) [triage] [usecase]

acoburn: and w3c/lws-ucs#65

<gb> Issue 65 [UC] Context-based Access and Usage Control (by termontwouter) [triage] [usecase]

acoburn: and w3c/lws-ucs#63

<gb> Issue 63 [UC] Policy Management for Derived Resources (by termontwouter) [triage] [usecase]

acoburn: and w3c/lws-ucs#61

<gb> Issue 61 [UC] Intuitive (Data Type-Based) Policy Management (by termontwouter) [triage] [usecase]

<Zakim> gibsonf, you wanted to ask what is a derived resource

<gibsonf1> gibsonf1

acoburn: the way I understand derived resource (63) is
… say you have a set of fixed resources, and then a set of resources that derive from that

gibsonf1: if you had a request to ???

<gibsonf1> a derived resource = derived resource

acoburn: it depend a lot on how resources are defined in some

<gibsonf1> question was: Wouldn't request to a derived resource be no different than a request to any resource with respect to authoriazation?

acoburn: other issue that's related is w3c/lws-ucs#60

<gb> Issue 60 [UC] Freedom of policy modelling language/mechanism (by termontwouter) [triage] [usecase]

acoburn: the UC also introduces more complexity
… yet another is w3c/lws-ucs#89

<gb> Issue 89 [UC] Share data with a (potentially large) community/group of people (by mrkvon) [triage] [usecase]

acoburn: finally w3c/lws-ucs#64

<gb> Issue 64 [UC] Fine-Grained Control Over Resource Access (by termontwouter) [triage] [usecase]

acoburn: Last thing I want to mention

<dmitriz> what I find fascinating about these issues, is that they highlight the central tension in this WG.

acoburn: 1 approach that is in Solid is that access control is internal
… it's fairly constrained bu it's sinternal to the sercer

<dmitriz> is our goal comprehensive documentation of use cases? (in which case, this is pretty great). Is our goal shipping a first iteration spec that we can build upon? In that case, most of these are _way_ too advanced / ambitious

acoburn: 1 approach would be to handling this externally to the server
… e.g. defining what an access token would look like

<Zakim> pchampin, you wanted to ask about what 'external to the server' means

acoburn: that would allow other kind of abilities

pchampin: not sure what you mean by external to the server
… it could be that the server relies on an external server for access control mechanism
… it may be an implementation detail not in scope for this group
… could you clarify exactly what you mean, acoburn

acoburn: this has impact on what we do for the protocol
… is the LWS protocol going to define what's in an identity token
… one approach would be to define a format of an access token
… separately, you could define how you get access to the token
… using a binding for lots of implementations
… that what I ment by external

<Zakim> gibsonf, you wanted to weigh in on implentation effects

gibsonf1: wtith RDF on the server we cash things
… we use the authorisation with cashing mechanisms
… it may be very difficult to handle this at scale

uvdsl: can you be more specific wrt the scope of the protocol

acoburn: authorization is very much in scope of the protocol
… but how we scope it?
… are we going to scope it to a particular way of getting authorization?

<ericP> i think "auth out of scope" means "auth *control* is out of scope", i.e. the mechanism for setting authorizations won't be portable between systems

acoburn: if we define one thing, how we compare to another way

uvdsl: there are 2 questions, do we choose a specific policy solution?
… and if the authorization server and the resource server are distinct, what alternative to them being coupled ?
… I am thinking at how it work with Solid and how this is handled in different ways

acoburn: the 2 entities could coexist may do not necesssarily have to

<Zakim> gibsonf, you wanted to ask what are the things WAC just cant do?

gibsonf1: you may want to couple the 2 if in your pod some data is used for authorization
… if decoupled, it would be more difficult to understand what's going on in the pod
… what WAC cannot do that other protocols can do?

acoburn: one answer is WAC can do anything you want
… but maybe undermining interoperability
… it cannot do time based access control
… on its own
… so you need something on top that's not interoperable
… or you can't do consent based access by itself
… also with 1000000 of users,
… things start to slow down much
… from the top of my head, and then more things

gibsonf1: you could have something like one emply puts the WebId and that works for the whole company employees

acoburn: imagine you need the doctor to access one's health data
… the doctor could delegate access
… but the others cannot delegate again

Beau: we talked about WAC and access control policy
… WAC can define groups and user agents but cannot define fine grained policies
… you can do it with a policy language

<Beau> https://solid.github.io/authorization-panel/acp-specification/

Beau: Solid can use ACP instead of WAC

acoburn: and there are things ACP can't do and WAC can

uvdsl: these conversations about WAC can and can't do happened already in the Solid CG
… in fact some things re. ACP are not done by ACP per se but by Inrupt's implementation

acoburn: having a particular policy language is one appraoch
… there are other appraoches
… what are the tradeoffs of all the appraoches

<uvdsl> To add what I wanted to raise for the protocol: If we acknowledge that WAC is incomplete, why are we not considering to extend/modfiy WAC?

Minutes manually created (not a transcript), formatted by scribe.perl version 244 (Thu Feb 27 01:23:09 2025 UTC).

Diagnostics

Succeeded: s/???/Monsecom

Succeeded: s/noboday/nobody

Succeeded: s/bacl/back

Succeeded: s/???/a derived resource

Succeeded: s/Beau: you could/gibsonf1: you could/

Maybe present: Monsecom, RazaN

All speakers: acoburn, Beau, gibsonf1, Monsecom, pchampin, RazaN, uvdsl

Active on IRC: acoburn, AZ, Beau, bendm, cpn, dmitriz, eBremer, ericP, gibsonf1, kaefer3000, Monsecom, pchampin, RazaN, uvdsl