Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • May 20 16:09
    justinwb assigned #408
  • May 20 09:41

    csarven on main

    Update solid-oidc reference. Ba… (compare)

  • May 19 14:37
    acoburn synchronize #408
  • May 19 01:36
    acoburn synchronize #408
  • May 18 16:23

    csarven on main

    Add 2022-05-18 minutes (compare)

  • May 18 12:33
    matthieubosquet synchronize #408
  • May 18 10:07
    matthieubosquet synchronize #408
  • May 18 10:00
    matthieubosquet synchronize #408
  • May 18 09:45
    matthieubosquet synchronize #408
  • May 18 09:34
    matthieubosquet synchronize #408
  • May 18 01:55
    matthieubosquet commented #408
  • May 18 01:51
    matthieubosquet synchronize #408
  • May 14 12:25
    acoburn edited #408
  • May 14 12:17
    acoburn synchronize #408
  • May 13 19:56
    acoburn opened #408
  • May 13 16:13

    csarven on main

    Add privacy-principles (compare)

  • May 12 17:03
    csarven commented #407
  • May 12 17:02
    csarven commented #407
  • May 12 16:39
    csarven review_requested #407
  • May 12 16:39
    csarven opened #407
Martynas Jusevicius
@namedgraph_twitter
no
foaf:Agent is used if no WebID client cert is present

based on the WAC https://www.w3.org/wiki/WebAccessControl

Servers are required to recognize the class foaf:Agent as the class of all agents. This indicates that the given access is public. In some cases this will mean that authentication is therefore not required, and may be skipped. When a resource is being written, however, it may be necessary to associate the change with some kind of ID for accountability purposes.

Aaron Coburn
@acoburn
acl:AuthenticatedAgent makes sense if the server constrains the identity providers that are trusted. In traditional web applications with external identity providers, this tends to be a closed set (e.g. login with Google, FB and Twitter but not any arbitrary identity system), and there, acl:AuthenticatedAgent makes sense. If, however, identity in Solid is entirely decentralized and if resource servers need to be able to negotiate auth tokens from an open set of identity providers, then there is effectively no difference between an “authenticated” agent and an unauthenticated agent: I can create an ephemeral identity provider with an ephemeral WebID and call that an acl:AuthenticatedAgentwhich could get a higher level of access just because I have some scripts running this ephemeral infrastructure
Sarven Capadisli
@csarven
It seems that the current definition of acl:AuthenticatedAgent is more like IdentifiedAgent. For AuthenticatedAgent to be truly authenticated, server needs to be able to verify the authenticity of the identity claim.
Martynas Jusevicius
@namedgraph_twitter
well we only support WebID-TLS
so the presence of a valid WebID cert is the verification
i think the class name is appropriate
what’s the difference between “identification” and “authentication” really?
Sarven Capadisli
@csarven
Generally: claiming and verifying
Dmitri Zagidulin
@dmitrizagidulin
given how trivial it is to generate an ephemeral web id, I do think the AuthenticatedAgent term is useless.
Justin Bingham
@justinwb
+1
Martynas Jusevicius
@namedgraph_twitter
@csarven WebID-TLS does both
@dmitrizagidulin not sure what you mean
with WebID-TLS, you need to generate a certificate and host the WebID profile somewhere
that requires some effort at least
and you know which domain it is coming from
Martynas Jusevicius
@namedgraph_twitter

So W3C (Sir Tim, no less) turned down the opportunity to host the JS standard because they felt the web should be declarative-only. Imagine a declarative-only web today.

"Javascript: the first 20 years" https://buff.ly/30TpkPu (Brilliant article in a brilliant journal!)

https://twitter.com/technosophos/status/1273614442874445825?s=20

Sarven Capadisli
@csarven
Are they cases where either one (or both) would be preferable for servers to share request error details with clients: i) in response body ii) link relation using ldp:constrainedBy with target URI identifying the constraints?
Dmitri Zagidulin
@dmitrizagidulin
does ldp:constrainedBy have anything to do with errors?
Sarven Capadisli
@csarven
Server is supposed to include that header in 4xx responses.
Dmitri Zagidulin
@dmitrizagidulin
oh really? is that from LDP?
Sarven Capadisli
@csarven
LDP doesn't specify the details for the representation. In fact, it doesn't even need to be in RDF. In Solid, I think we are leaning on having the details in RDF.
Dmitri Zagidulin
@dmitrizagidulin
q+
Dmitri Zagidulin
@dmitrizagidulin
q-
Sarven Capadisli
@csarven
Sarven Capadisli
@csarven
I think this PR on how constrainedBy can be used in Solid is a good start solid/specification#185 . If there is nothing out of the ordinary for the rough consensus, can we get that merged?
For Problem Details (client error), we can do another PR once we have more specifics on the message schema.
Dmitri Zagidulin
@dmitrizagidulin
@csarven what's the pain point that #185 solves?
Sarven Capadisli
@csarven
It was for issue solid/specification#44
Sarven Capadisli
@csarven
I think it would fall under the UC here: https://www.w3.org/TR/ldp-ucr/#dfn-uc1 -- access guidance, https://www.w3.org/TR/ldp-ucr/#dfn-nf1.1
Dmitri Zagidulin
@dmitrizagidulin
@csarven I guess my main question is, why is it a MUST?
(given that we don't have need for this yet, don't have vocab to define how to constrain or what to constrain, etc)
Sarven Capadisli
@csarven
@dmitrizagidulin For the constraints defined in the Solid spec, that's not much of an issue because whatever is defined in the spec will be common to all servers and clients in the ecosystem.
In the case of a response body describing the problem details, yes, it;d be necessary provide the vocab. That's covered in solid/specification#28
Sarven Capadisli
@csarven
Even just making it possible to describe the problem details can help clients without necessarily having prior knowledge of the vocab.
Dmitri Zagidulin
@dmitrizagidulin
I think what might help me is an example
because as an implementer, I'm looking at that spec text, and it says a server MUST advertise constraints, and I'm thinking.. what constraints? In what cases? and how do I express them?
Sarven Capadisli
@csarven
Say server denies an update. It can return 409 and link to the spec constraint saying that server doesn't allow client to update containment triples.
Dmitri Zagidulin
@dmitrizagidulin
so what's an example of that?
(so that I know what to return on a 409)
Sarven Capadisli
@csarven
With this specific example, it'd be https://www.w3.org/TR/ldp/#ldpc-put-mbrprops
(or whatever the URI is in the Solid spec)
Dmitri Zagidulin
@dmitrizagidulin
I see, thanks
Justin Bingham
@justinwb
@csarven i had commented on the ticket but do you assume shape validation falls into this bucket
Sarven Capadisli
@csarven
Not sure about exact categorisation. There are similar concepts in play but I'd say that SV is possibly a specific kind of a constraint (ie. linked via constrainedBy). Besides that, constrainedBy can handle HTTP, authn/z based constraints, in addition to data.
Comes back around to what I've asked several times about the relationship/diff with solid:shape...
Justin Bingham
@justinwb
@csarven are you good for session today?
Sarven Capadisli
@csarven
Yea.