solid/data-interoperability-panel

2021-12-07 Agenda: https://hackmd.io/iVlburEVSRmhFxAtjtmAsw | Code of Conduct: https://github.com/solid/specification#code-of-conduct

elf Pavlik

@elf-pavlik

Eric Prud'hommeaux

@ericprud

if we invent an Organization that subclasses Agent, maybe we should come up with some name for non-org Agents so we don't have the same PITA in LDP where there's no way to say something is a non-Container Resource

elf Pavlik

@elf-pavlik

https://www.w3.org/TR/vcard-rdf/#d4e2121

elf Pavlik

@elf-pavlik

q+ connecting to granting access based on remote data registration

elf-pavlik @elf-pavlik will need to go in 10 min

elf-pavlik @elf-pavlik has to go for call with NLnet

Justin Bingham

@justinwb

@/all PR with minutes from today - solid/data-interoperability-panel#106

elf Pavlik

@elf-pavlik

for anyone editing sequence diagrams and using vscode: https://marketplace.visualstudio.com/items?itemName=tomoyukim.vscode-mermaid-editor
i'm working on updates to #97

elf Pavlik

@elf-pavlik

https://github.com/solid/data-interoperability-panel/pull/97#issuecomment-834692289 now includes a second diagram, it would be great to receive some feedback before our meeting tomorrow

Justin Bingham

@justinwb

@/all HackMD for today - https://hackmd.io/@hAlm_W2-T-q16baquNyRXg/S1Y-_Wdu_

Justin Bingham

@justinwb

sorry wrapping prior call - need 2m :slight_smile:

Justin Bingham

@justinwb

solid/data-interoperability-panel#97

hRicklefs

@hRicklefs

https://plantuml.com/

elf Pavlik

@elf-pavlik

https://github.com/solid/authentication-panel/issues/145#issuecomment-837067449

for anyone editing sequence diagrams and using vscode: https://marketplace.visualstudio.com/items?itemName=tomoyukim.vscode-mermaid-editor

elf Pavlik

@elf-pavlik

q+ re VC formats

https://www.w3.org/TR/vc-data-model/#json-web-token

Justin Bingham

@justinwb

https://solidproject.org/TR/protocol

elf Pavlik

@elf-pavlik

solid/specification#177

Justin Bingham

@justinwb

shapetrees/specification#44

elf Pavlik

@elf-pavlik

q+ re instance

preview https://deploy-preview-44--shapetrees-specification.netlify.app/

elf Pavlik

@elf-pavlik

@justinwb I think we should change default branch to main and prune branches in the repo https://github.com/solid/data-interoperability-panel/branches

I could help with this chore, I would just need higher permission on the repo.

Actually I have rights to delete branches from merged PRs so I did it. You still have some old branches which never got PR https://github.com/solid/data-interoperability-panel/branches/all

elf-pavlik @elf-pavlik just realized that HackMD has build in mermaid diagrams support https://hackmd.io/YB1QIm9lS3OEVW5KUFDz2g?view

elf Pavlik

@elf-pavlik

@justinwb @ericprud I've been looking at Data Grant and it's flavors Referenced Data Grant, Remote Data Grant, Referenced Remote Data Grant

As I understand it, depending on value of interop:scopeOfGrant one of other properties is expected to be set on the Grant itself.

For example interop:scopeOfGrant interop:SelectedInstances ; would require setting hasDataInstance

Now looking at shape

<#DataGrantShape> {
  a [ interop:DataGrant ] ;
  interop:hasAccessGrant @<#:AccessGrantShape> ;
  interop:satisfiesAccessNeed @<#:AccessNeedShape> ;
  interop:registeredShapeTree IRI ;
  interop:hasDataRegistration IRI ;  
  interop:accessMode @<#:AccessModes>+ ;
  interop:scopeOfGrant @<#:DataGrantScopes> ;  
  interop:hasDataInstance IRI* ;  
  interop:hasReferencedDataGrant @<#:ReferencedDataGrantShape> 
}

zero or more hasDataInstance are valid, no matter value of scopeOfGrant

Would it make sense to create more specific shapes (one per data grant scope), which would have stricter definitions? eg.

<#DataGrantShapeSelectedInstances> {
  a [ interop:DataGrant ] ;
  interop:hasAccessGrant @<#:AccessGrantShape> ;
  interop:satisfiesAccessNeed @<#:AccessNeedShape> ;
  interop:registeredShapeTree IRI ;
  interop:hasDataRegistration IRI ;  
  interop:accessMode @<#:AccessModes>+ ;
  interop:scopeOfGrant [  interop:SelectedInstances ] ;  
  interop:hasDataInstance IRI+ ;  
  interop:hasReferencedDataGrant @<#:ReferencedDataGrantShape> 
}

etc.

elf Pavlik

@elf-pavlik

I'm afraid that with just one liberal shape, invalid data grants would still validate.

elf Pavlik

@elf-pavlik

I would even consider defining specific sub classes of DataGrant instead of using interop:scopeOfGrant.

<#DataGrantForSelectedInstancesShape> {
  a [ interop:DataGrantForSelectedInstances ] ;
  interop:hasAccessGrant @<#:AccessGrantShape> ;
  interop:satisfiesAccessNeed @<#:AccessNeedShape> ;
  interop:registeredShapeTree IRI ;
  interop:hasDataRegistration IRI ;  
  interop:accessMode @<#:AccessModes>+ ;
  interop:hasDataInstance IRI+ ;  
  interop:hasReferencedDataGrant @<#:ReferencedDataGrantShape> 
}

BTW while referencing specific instances why do we also reference the Data Registration? Grantee won't be able to list all the instances, so I assume it won't even have read access to that Data Registration. Possibly we would just have

<#DataGrantForSelectedInstancesShape> {
  a [ interop:DataGrantForSelectedInstances ] ;
  interop:hasAccessGrant @<#:AccessGrantShape> ;
  interop:satisfiesAccessNeed @<#:AccessNeedShape> ;
  interop:registeredShapeTree IRI ;
  interop:accessMode @<#:AccessModes>+ ;
  interop:hasDataInstance IRI+ ;  
  interop:hasReferencedDataGrant @<#:ReferencedDataGrantShape> 
}

elf Pavlik

@elf-pavlik

I captured above in solid/data-interoperability-panel#109 which now would be preferable place to follow up.

elf Pavlik

@elf-pavlik

@justinwb when someone would create ReferencedDataGrant with scope AllInstancesfor DataRegistration xyz, rather thatn creating regular Data Grant for xyz?
I understand ReferencedDataGrant with scope InheritInstances since it relies on which instances are selected in "parent" grant. Actually I don't see reason to use ReferencedDataGrant in any other case other than with InheritInstances.

Justin Bingham

@justinwb

@elf-pavlik yeah i think would could probably tighten the scope of those shapes up more

they were a bit fluid as we kept optimizing the grant structure which is why we probably had a more liberal schema

elf Pavlik

@elf-pavlik

@justinwb when someone would create ReferencedDataGrant with scope AllInstancesfor DataRegistration xyz, rather thatn creating regular Data Grant for xyz?
I understand ReferencedDataGrant with scope InheritInstances since it relies on which instances are selected in "parent" grant. Actually I don't see reason to use ReferencedDataGrant in any other case other than with InheritInstances.

@justinwb and :point_up:

elf Pavlik

@elf-pavlik

I'm trying to visualize different data grants:

AllInstances

InheritSelected

elf Pavlik

@elf-pavlik

SelectedInstances

green - can access
gray - doesn't even know about
red - has link but can't access

tomorrow i'll play with remote data registrations

elf Pavlik

@elf-pavlik

I wonder how others feel about changing Referenced (shape trees, data grants) to Related? This would better fit established concepts like relational database, entity-relationships etc.
I know naming things is hard! I have impression that in this case someone used st:references in Shape Trees spec and it bubbled up from there.

Justin Bingham

@justinwb

yeah that’s exactly where it came from (that it was represented a reference / virtual tree)

main issue i would have with related is that there is still an implied (albeit) virtual hierarchy and i don’t know that related conveys that well enough.

Justin Bingham

@justinwb

@justinwb when someone would create ReferencedDataGrant with scope AllInstancesfor DataRegistration xyz, rather thatn creating regular Data Grant for xyz?
I understand ReferencedDataGrant with scope InheritInstances since it relies on which instances are selected in "parent" grant. Actually I don't see reason to use ReferencedDataGrant in any other case other than with InheritInstances.

InheritInstances is definitely the most important case, but even when presenting AllInstances, there’s a difference in how a root (e.g. project) grant and the behavior of types (like tasks) that inherit based on the root being selected.

we should walk through some examples on tuesday that i’ve got broken out in turtle

elf Pavlik

@elf-pavlik

Sounds great @justinwb. Let's take a look at your examples tomorrow. I'll start HackMD with agenda

pad for tomorrow: https://hackmd.io/17e4EC9CTUGRwA6fR_-5LQ

Where communities thrive

People

Repo info

Activity