W3C

– DRAFT –
Linked Web Storage

12 January 2026

Attendees

Present
acoburn, Beau, bendm, eBremer, ericP, gibsonf1, laurens, Luke, pchampin, ryey, TallTed, uvdsl
Regrets
-
Chair
laurens
Scribe
acoburn, laurens

Meeting minutes

Introductions & Announcements

matthieu: currently work at ODI, worked on some solid specs

Initial CRUD with proposed metadata handling (PR #37 )

laurens: PR by ebremer. Has been discussed since mid-Nov

eBremer: since last meeting, rebasing to make reviewing easier

<Zakim> gibsonf, you wanted to ask about subject issue for 37

gibsonf: wrote about the use of subject resources and potential changes
… simple way uses a query string with subject URI

<gibsonf1> w3c/lws-protocol#37 (comment)

<gb> Pull Request 37 Initial CRUD with proposed metadata handling (by ebremer)

tallted: I have not been able to review, some more time would be nice

laurens: should we postpone to next meeting

ebremer: Fred's suggestion relates to query

gibsonf1: this is not about query, it's about CRUD operations on subject resources
… esp subject resources that are outside the domain of a storage
… idea is to qualify reads and writes to a particular URI

laurens: I see use cases, but shouldn't be a separate use case. This may be more complicated than introducing a query param. Impact may be broader
… also relates to authZ/authN

gibsonf1: this has big implications for interop

pchampin: we seem not to be talking about the same thing
… fred talks about resources in the broad (RDF) sense
… e.g. subject URIs that are outside the domain of a storage
… however, the proposed changes are not necessary
… the resources described in this PR are Information Resources contained in a storage
… this allows RDF files to be added to a storage with any subject
… thus, retrieval of a subject resource requires some form of query

gibsonf1: one can put arbitrary triples in files in a storage

<pchampin> by "file" I meant "information resource": some content that I can retrieve by dereferncing a the URI; I'm not assuming it is stored as a file on the server

gibsonf1: but this is unhelpful if I want to give access to a ??? resource to someone else

laurens: LWS is defining crud operations on information resources, I fail to see how a LWS server could display information resources that are out of domain of a storage
… I fail to see how adding query params would fix this

gibsonf1: in terms of files, these subject resources can be anywhere, but how do I find this information?

laurens: "how do I find something" is query
… and discovery

gibsonf1: an information resource may be prefixed with an LWS URI
… this would be very limited

luke: would like clarification -- how does the linking work?

gibsonf1: I would like to be able to have something more than just a google drive-like interface

luke: this seems like something else

gibson: the response provides the triples with a particular subject URI

<Zakim> acoburn, you wanted to talk about scope of LWS

acoburn: I think we're talking about a couple of different layers
… any implementation that uses specifications is going to combine many different specs.
… A specification isn't going to encompass the entire implementation.
… LWS tries to be really specific about that low level of information resource and CRUD on that resource.
… The content of that resource is a different layer.
… The content could be RDF, JSON, XML, ...
… In Fred's case he could use another specification on top of LWS, like SPARQL.
… This could result in the merging of a triple store and LWS server.
… If SPARQL isn't the right thing, QPF or TPF could be the solution perhaps.
… In LWS we want to define the low level information resource.

<Zakim> gibsonf, you wanted to respond

gibsonf1: not talking about the content, rather the address

<dmitriz> I think it might help to use example URLs in this discussion..

tallted: if I'm writing to my storage, there will be a location of the storage, represented by a URI
… there are objects in that storage, e.g., documents or something that resembles a database
… a filesystem can be conceived as a database of a kind
… the issue EricP was discussing related to misunderstanding of the technologies at hand
… i.e. a preceived need to give every protein a localized URI
… in that case, a relative uri was used. Using those different identifiers for the same protein can be useful
… each entity is different, but each refers to the same protein. You can use owl:sameAs to make use of this

tallted: using sameAs allows us to talk about many different entities as if they are the same

tallted: this is like referring to Ted using various different identifiers, all of which refer to the same individual
… having the multiple names are helpful because I can then list all the names I want, excluding those I don't want
… understand Fred's use case, this seems to be asking for a query that includes the entity identifier and get the information from
… all the resources the include that subject resource
… what you're asking for is the magic for how those subject identifiers are all associated

<gb> Issue 215 [UC] Mechanism to access a resource identifier in a storage with identifier that points to a different storage (by gibsonf1) [triage] [usecase]

tallted: In that case, Fred needs to know all the places where a particular subject resource is referenced
… in order for that to work, if the software needs to do that, it is very complicated for software. It will require a lot of computational
… resources
… realistically, it becomes intractable

<Zakim> gibsonf, you wanted to talk about sameas

gibsonf: I'm talking about anyone who stores information with a subject URI
… i.e. a scientist can request all examples with a common subject URI

tallted: Fred is talking about a limited application of lookup magic. I.e., given a URI, go to N storages and merge the data
… this is a small kind of magic, but it is still magic
… this is the general idea of linked web storage, but there is a lot of magic involved
… we have discussed this for several weeks but haven't made headway

eBremer: I believe I am following what Fred is saying, and we have similar use cases
… e.g., different users need access to different resources & URIs
… would like to talk further

<dmitriz> could we do a special topic call on this?

<dmitriz> haha true

<pchampin> +1 to separate the "in scope for CRUD" and "in scope for LWS"

laurens: we could decide whether it is in or out of scope of CRUD

<dmitriz> -1 (out of scope with CRUD, in scope with a separate query feature)

STRAW POLL: is Fred's feature in scope for CRUD?

<gibsonf1> +1

<acoburn> -1

<matthieubosquet> -1

<pchampin> -1

<Luke> -1

<laurens> -1

<eBremer> -1

<ryey> -1

<TallTed> -1 (out of scope for CRUD; in scope for LWS; will be implementable by chunks of CRUD and other)

<ericP> -1 but willing to consider a detailed technical write-up

<laurens> w3c/lws-ucs#215

<gb> Issue 215 [UC] Mechanism to access a resource identifier in a storage with identifier that points to a different storage (by gibsonf1) [triage] [usecase]

laurens: most see this as out of scope for CRUD, we can discuss further in the open issue
… we will try to resolve PR 37 to next week. Please review by next week

Storage Description Resource (PR #53 )

acoburn: This is a PR that has been open for a couple of weeks.
… Shortly before this meeting only ryey had comments.
… I don't think we can vote on this today.
… This is a way of defining a particular resource in a storage that defines capabilities and extension points.
… For example if the storage has a notification service, SPARQL endpoint, support for binary patch, ...
… It is a way of adding capability and endpoint discovery to storage in an interoperable way.
… Are there other comments? I assume we cannot proceed to a vote today.

dmitriz: Why add one more layer of indirection for the capabilities

acoburn: Why not have both?
… A lot of linked data systems are already pretty chatty.
… But I am not opposed to having a storage description have a linkset.

dmitriz: I think that is exactly the reason to have a linkset.

acoburn: dmitriz would you be able to write up in the PR what that might look like?

<Zakim> pchampin, you wanted to ask what would that level of indirection buy us

pchampin: what problem does this indirection solve?

pchampin: I don't really see what this level of indirection will get us.
… perhaps the binding between the data model and the specific media type can be made more explicit

gibsonf1: is there a unique URI for a storage?

<Zakim> gibsonf, you wanted to ask why not just have a uri for the storage itself to use to get information about the storage itself?

acoburn: Historically the root of a storage (the storage URI) becomes a container.
… If you want to give access to that root container, you are perhaps giving broad access to the storage.
… With this resource the intent is for it to be public.
… It gives details pertaining to how you might request access.
… Its authorization is different to the root of the storage.

gibsonf1: conflating a root container with a storage URI causes trouble

<Zakim> gibsonf, you wanted to ask to not conflate a root container with the LWS storage uri

pchampin: the example seems to imply that the URI suggests a root container, but there is nothing normatively that requires this
… also, there is nothing that prevents the description being the same as the root URI

<gibsonf1> why would storage desciprion uri be different that storage uri

<pchampin> gibsonf1 I agree that it might not be necessary in most cases... the storage description is a "representation" (in REST terms) of the storage

laurens: we will continue discussing Fred's use case in the related issue
… we can close off for today

Minutes manually created (not a transcript), formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).

Diagnostics

Succeeded: s/RRSAgent/

Succeeded: s/own:sameAs/owl:sameAs/

Maybe present: dmitriz, gibson, gibsonf, matthieu

All speakers: acoburn, dmitriz, eBremer, gibson, gibsonf, gibsonf1, laurens, luke, matthieu, pchampin, tallted

Active on IRC: acoburn, Beau, bendm, dmitriz, eBremer, ericP, gibsonf1, laurens, Luke, matthieubosquet, pchampin, ryey, TallTed, uvdsl