Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add notion of "owner" and requirements #264

Merged
merged 2 commits into from
Jun 16, 2021
Merged

Add notion of "owner" and requirements #264

merged 2 commits into from
Jun 16, 2021

Conversation

csarven
Copy link
Member

@csarven csarven commented May 19, 2021

Hello Reviewers,

In response to the following (and possibly other) issues on the notion of "owner":

If describing the relationship between Solid data pod / storage "owner" and AWWW's "URI ownership" is required, it can be done in another PR.

On merge of this PR, a follow-up PR will be made to the https://github.com/solid/vocab/ repository requesting to add and define the term "owner" in the Solid Terms vocabulary: http://www.w3.org/ns/solid/terms#owner. Proposal (derived from http://www.w3.org/ns/auth/acl#owner ):

solid:owner
    a rdf:Property ;
    dc:issued "2021-05-19"^^xsd:date ;
    rdfs:comment "The person or other agent which owns the data storage. For example, the owner has implicit control of a data storage."@en ;
    rdfs:isDefinedBy <http://www.w3.org/ns/solid/terms#> ;
    rdfs:range foaf:Agent ;
    rdfs:label "owner"@en .

Copy link
Contributor

@timbl timbl left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks good to me

@bblfish
Copy link
Member

bblfish commented May 19, 2021

The domain of the relation is not defined explicit. I guess it is a Pod, or would it be a ldp:Container?

@csarven
Copy link
Member Author

csarven commented May 19, 2021

Intentionally undefined for the time being. Implementation feedback would help. In the Protocol, servers exposing the storage resource (root container) also has a rel="type" link relation http://www.w3.org/ns/pim/space#Storage.

If we lean towards pim/space, it could also be pim:ControlledStorage. The property could alternatively go under pim/space eg. pim:owner.

Copy link

@kidehen kidehen left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Intentionally undefined for the time being. Implementation feedback would help. In the Protocol, servers exposing the storage resource (root container) also has a rel="type" link relation http://www.w3.org/ns/pim/space#Storage.

If we lean towards pim/space, it could also be pim:ControlledStorage. The property could alternatively go under pim/space eg. pim:owner.

You should include the rdf:domain for clarity.
As I see it, the rdf:domain would be rdfs:Resource if want to be broad.

solid:owner
    a rdf:Property ;
    dc:issued "2021-05-19"^^xsd:date ;
    rdfs:comment "The person or other agent which owns the data storage. For example, the owner has implicit control of a data storage."@en ;
    rdfs:isDefinedBy <http://www.w3.org/ns/solid/terms#> ;
    ## Property Domain Addition
    rdfs:domain rdfs:Resource ;
    rdfs:range foaf:Agent ;
    rdfs:label "owner"@en .

@kidehen
Copy link

kidehen commented May 19, 2021

Intentionally undefined for the time being. Implementation feedback would help. In the Protocol, servers exposing the storage resource (root container) also has a rel="type" link relation http://www.w3.org/ns/pim/space#Storage.

If we lean towards pim/space, it could also be pim:ControlledStorage. The property could alternatively go under pim/space eg. pim:owner.

You should include the rdf:domain for clarity.
As I see it, the rdf:domain would be rdfs:Resource if want to be broad.

solid:owner
    a rdf:Property ;
    dc:issued "2021-05-19"^^xsd:date ;
    rdfs:comment "The person or other agent which owns the data storage. For example, the owner has implicit control of a data storage."@en ;
    rdfs:isDefinedBy <http://www.w3.org/ns/solid/terms#> ;
    ## Property Domain Addition
    rdfs:domain rdfs:Resource ;
    rdfs:range foaf:Agent ;
    rdfs:label "owner"@en .


<p>When a server wants to advertise the owner of a storage, the server MUST include the <code>Link</code> header with <code>rel="http://www.w3.org/ns/solid/terms#owner"</code> targeting the URI of the owner in the response of HTTP <code>HEAD</code> or <code>GET</code> requests targeting the root container.</p>

<p class="note" role="note"><span>Note</span>: When a server supports multiple storages, there must be complete trust between its owners.</p>
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
<p class="note" role="note"><span>Note</span>: When a server supports multiple storages, there must be complete trust between its owners.</p>
<p class="note" role="note"><span>Note</span>: When a server supports multiple owners, there must be complete trust between its owners.</p>

?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think when a storage has multiple owners, trust between them is generally expected or well-understood. This is really about when a server supports multiple storages with different owners under the same host.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure I fully understand the relationship/geometry between server, owner and storage (and Pod).

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How does it apply to https://solidcommunity.net/ ? One server supports multiple users but uses subdomain for each storage.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is really about when a server supports multiple storages with different owners under the same host.

A Pod hosting scenario (per @elf-pavlik's point) would seem to fall under this category where a given server has multiple storages, with each having different owners, who may not know/trust each other.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

?

The key word that I used was "host" but perhaps I should've used "scheme, host, and port" to be more accurate. This is boils down to origin (see RFC 6454). solidcommunity.net designates a unique "scheme, host, and port" with one "storage" to one "owner". It is currently a non-issue for solidcommunity.net as I understand it.

The main point remains in that if a service (like solidcommunity.net) were to allow multiple storages under the same "scheme, host, and port", then the owners of the storages (which could be the same owner) need to establish "trust" among themselves.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pod hosting isn't necessarily exclusive to sub-domains for segregation. Pod hosting scenarios can, and do, use path based segregation (e.g. https://host.example/pod-one, https://host.example/pod-two). In cases like those, the scheme, host, and port are the same, but still you have the likely scenario of different hosted pods owned by social agents that do know know and/or trust each other.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I responded to the original inquiry about subdomain based services and explained why that was a non-issue. Yes, multiple storages (on the same origin) with different owners has been possible for long time. It was never prohibited to begin with! (Heck, we've had different types of storages as well as workspaces... ) One of the reasons for the whole set of requirements around storage was acknowledging that reality (see PR #208 and https://solidproject.org/TR/protocol#storage ).. making sure non-overlapping URI ((path) space are used, among other things.

Trivial to get a wholesale scheme-host-port blocked (rendering it useless) by a service because of one bad actor, and hard to isolate the bad actors in that system. Not to forget CORS considerations.

The note doesn't prohibit servers/services that possibility. Both the service provider and the owners of the storages should be aware of what it entails.

The point in the note still stands and relevant for any set up.

protocol.html Outdated Show resolved Hide resolved
@csarven
Copy link
Member Author

csarven commented May 19, 2021

@kidehen No objection to including rdfs:Resource but I don't think it is needed. Wouldn't that always be inferred?

@kidehen

This comment has been minimized.

protocol.html Outdated Show resolved Hide resolved
protocol.html Outdated Show resolved Hide resolved
@RubenVerborgh
Copy link
Contributor

    rdfs:comment "The person or other agent which owns the data storage. For example, the owner has implicit control of a data storage."@en ;

The "for example" doesn't seem to work well.

@bblfish
Copy link
Member

bblfish commented May 19, 2021

So if I understand correctly in N3 we have

# we first define the transitive closure of ldp:contains
:contains٭ owl:TransitiveProperty;
     rdfs:subPropertyOf ldp:contains .

# we then have the N3 rule
{ ?rootLdpc :owner ?agent } => {  
     [] acl:agent ?agent ;
        acl:mode acl:Control;
        acl:accessToClass [ a owl:Restriction;
              owl:onProperty [ owl:inverseOf :contains٭ ];
              owl:hasValue ?rootLdpc 
         ]
}

If it were a little clearer what the domain of the relation was (the domain-name perhaps?) one could adapt the definition.

@kjetilk
Copy link
Member

kjetilk commented May 20, 2021

Your proposal to define it in terms of pim:Storage is better, but still seems somewhat vague then. It leaves open whether we are identifying electronic devices via a hardware serial number, Hardware UUID, Provisioning UDID (info I find available for my computer) or with a domain. Are we interested in the ownership of the hardware, or in the topological position of the device on the internet? For the latter, it could be a MAC address or a domain name.

Alright, perhaps we should have a solid:Storage and tighten the definition appropriately?

I agree that we should have the domain in there as it very clearly binds the definition of an owner to the definition of a storage, which is very clearly specified in terms of the URI hierarchy in the protocol spec, which is again a rigorous definition AFAICS. It is therefore important to have a RDFS class that is equally strongly tied to the protocol's definition of storage. I can see that the pim:Storage definition can be interpreted rather broadly, so I say we tighten it.

protocol.html Show resolved Hide resolved
@kjetilk kjetilk linked an issue May 25, 2021 that may be closed by this pull request
protocol.html Show resolved Hide resolved
protocol.html Show resolved Hide resolved

<p>When a server wants to advertise the owner of a storage, the server MUST include the <code>Link</code> header with <code>rel="http://www.w3.org/ns/solid/terms#owner"</code> targeting the URI of the owner in the response of HTTP <code>HEAD</code> or <code>GET</code> requests targeting the root container.</p>

<p class="note" role="note"><span>Note</span>: When a server supports multiple storages, there must be complete trust between its owners.</p>
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is really about when a server supports multiple storages with different owners under the same host.

A Pod hosting scenario (per @elf-pavlik's point) would seem to fall under this category where a given server has multiple storages, with each having different owners, who may not know/trust each other.

@csarven
Copy link
Member Author

csarven commented May 31, 2021

@kidehen @bblfish @kjetilk :

The intended definition of solid:owner draws from acl:owner. It is broader in the sense that "there is a sense of right of control" over all data in a storage. Obviously no disagreement to make solid:owner more clear. Part of the reason why I left it vague was to work it out based on what's intended by the spec. The current rdfs:range foaf:Agent is intentional, so the term is definitely not about "agent owns/ownerOf thing".

We use the pim:storage property eg. agent has storage, where the range is defined to be an instance of pim:Storage. The arc from a storage to an agent is intended to be solid:owner. I think that's fine (ie. we can do rdfs:domain pim:Storage. Unless pim:Storage gets new subclasses (eg. for new devices/hardware..), I don't see what an equivalent like solid:Storage brings to the table.

@elf-pavlik

I borrowed/introduced "social entity" from AWWW's URI ownership. There are social expectations and processes. Possible to go from responsibility/accountability chain from data "ownership" to URI ownership.. domain name registration through IANA (another social entity). And as I've said at the beginning of this PR:

If describing the relationship between Solid data pod / storage "owner" and AWWW's "URI ownership" is required, it can be done in another PR.

AFAIK, "owner" in Solid is deemed to be a "social entity" of sorts, as opposed to a software agent. If there is existing discussion/work that incorporates non-living organisms under "owner", references would be great. (And to our future non-living organism overlords, I don't mean to discriminate, please go easy on us.)

@kjetilk
Copy link
Member

kjetilk commented May 31, 2021

I don't see what an equivalent like solid:Storage brings to the table.

Again, lets open a separate issue. I'd like to discuss tying it very explicitly to Solid's definition, which is rooted in its hierarchy, that's something that doesn't exist in the general case.

@csarven
Copy link
Member Author

csarven commented May 31, 2021

I'd like to discuss tying it very explicitly to Solid's definition, which is rooted in its hierarchy, that's something that doesn't exist in the general case.

Perhaps solid:Resource: #194 (comment)

Copy link
Member

@kjetilk kjetilk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have opened #265 and #266 to continue the discussion of the exposure of the owner relation and the domain of the relation there.

And given that these issues can be resolved there, I approve this PR.

Copy link
Contributor

@timbl timbl left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks good to me this is important for existing servers.

@csarven
Copy link
Member Author

csarven commented Jun 4, 2021

Implemented owner discovery in https://github.com/csarven/mayktso

curl -I https://linkedresearch.org/ldn/tests/

link: <http://www.w3.org/ns/pim/space#Storage>; rel="type"
link: <https://csarven.ca/#i>; rel="http://www.w3.org/ns/solid/terms#owner", <https://example.org/a43a1874-c51e-11eb-a893-3fad0d2fdf12#i>; rel="http://www.w3.org/ns/solid/terms#owner"

Copy link
Member

@justinwb justinwb left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  • I support the notion of a storage owner, because it's valuable for interoperability and data sharing workflows (e.g. where data in a given storage is shared by a non-owner with access to do so, and whose data they are sharing needs to be communicated)
  • I support public advertisement because there are cases where a non-privileged agent should be able to know whose storage they're accessing.

It is important that public advertisement can be anonymous, and I believe that most pod owners will choose to do so. This may impact the viability of certain workflows where discovery of ownership is necessary. That needs to be addressed, but doesn't have to block this submission, as we now have a separate thread where this is being discussed (#266). Similarly, I've also created #267 to continue this thread.

@csarven
Copy link
Member Author

csarven commented Jun 16, 2021

Was waiting for additional feedback from implementers before merging..

Mayktso currently only implements the discovery bit as mentioned earlier: #264 (comment)

Noting here that CSS is in the process of implementing this feature: CommunitySolidServer/CommunitySolidServer#720 .

NSS has a PR for it: nodeSolidServer/node-solid-server#1604

Solid-NextCloud has shown interest in implementing: pdsinterop/solid-nextcloud#48

Will update this comment with more.. or others (/me friendly nudges @kidehen @gibsonf1 @acoburn @bblfish .. ) can show/link to their implementations in the comments


As for the solid:owner definition, here is an update based on earlier discussion that can be PR'd to solid/vocab - we can finalise the exact definition there, instead of here:

solid:owner
    a rdf:Property ;
    dc:issued "2021-05-19"^^xsd:date ;
    rdfs:comment "The a person or a social entity that is considered to have control, rights and responsibilities of a data storage."@en ;
    rdfs:isDefinedBy <http://www.w3.org/ns/solid/terms#> ; # alternative: <https://solidproject.org/TR/protocol> ;
    rdfs:domain ws:Storage ;
    rdfs:range foaf:Agent ;
    rdfs:label "owner"@en .

As agreed, we'll work out related/orthogonal issues (that were deemed to be non-blocking).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
Status: Done
Development

Successfully merging this pull request may close these issues.

Should it be possible to find controller of storage via resources?