Specify activities for relevant storage operations

[elf-pavlik]

Based on the table in https://github.com/solid/web-access-control-spec/issues/85#issuecomment-913456115

Request	Operation	Access Modes	Level	Activity
GET /C/R	Read	acl:Read	Resource	* as:Read
HEAD /C/R	Read	acl:Read	Resource	* as:Read
OPTIONS /C/R	Read	acl:Read	Resource	* as:Read
POST /C/	Create + Update	acl:Append [1]	Resource, Content	? as:Add
PUT /C/R + If-None-Match	Create, Read	acl:Write, acl:Read [2]	Resource	? as:Create
PATCH /C/R + If-None-Match + INSERT	Create, Read	acl:Write, acl:Read [2]	Resource, Content	? as:Update
POST /C/R	Update	acl:Append (or acl:Write)	Content	? as:Add
PUT /C/R	Create?, Update	acl:Write [2][3]	Resource	? as:Create
PATCH /C/R + INSERT	Create?, Update	acl:Write [2][3]	Resource?, Content	as:Update
PATCH /C/R + DELETE + INSERT	Create?, Read, Update	acl:Write, acl:Read [2][3]	Resource?, Content [4]	as:Update
PATCH /C/R + DELETE	Update, Read	acl:Write, acl:Read	Content	as:Update
DELETE /C/	Delete	acl:Write, acl:Read	Resource	as:Remove
DELETE /C/R	Delete	acl:Write [5]	Resource	as:Delete

Also keeping in mind as a reference: https://communitysolidserver.github.io/CommunitySolidServer/6.x/usage/notifications/#notification-types

• All Read can be ignored
• Update results in as:Update (except updates to ldp:contains addressed by Create and Remove)
• Create only as:Add on the parent, since the newly created resource couldn't have been a topic of an existing channel
• Delete
  • as:Delete when the deleted resource was the topic
  • as:Remove when the parent container was the topic

@joachimvh in what case CSS sends as:Create notifications?

[CxRes]

As discussed in the meeting between me and @csarven, we would add an implementation guidance for as:Read, stating that origin servers need to ensure that they aggregate read (GET) events from an intermediary and broadcast them back to all intermediaries. Mechanisms for doing so would be outside the scope of the specification.

We need this because GET requests can be handled by HTTP intermediaries, whereas all write requests are always sent back to the origin server.

[CxRes]

Just be careful that as:Create and as:Add are distinct. @elf-pavlik Can you please add a fifth column, say "Activity", to the table that includes AS vocab mapping?

[elf-pavlik]

If you want to support as:Read I think there would need to be a way to opt-in for it since many (most?) use cases don't want to get notifications as:Read. There is also the question of who would be authorized to receive as:Read notifications. This is a privacy issue and probably only the storage owner would have access to such notifications.

[CxRes]

The context of this was audit functionality, so you are right this would be available to storage owner (or auditors who are given access to notifications but not contents/diffs). It is just that we do not want to shut off that possibility, even if we do not specify it right now. I think read will be allowed using a filter (just that it is "unfiltering").

[joachimvh]

@joachimvh in what case CSS sends as:Create notifications?

Create notifications get sent when a POST/PUT/PATCH request writes data to a resource that didn't exist yet. So the same situations where the required permissions would be different, as you require acl:Append permissions on a container when trying to create a new resource.

Besides that, Update gets sent when an existing resource is modified through PUT/PATCH, so not when it gets created initially, and Delete when it gets removed.

In the case of Create and Delete, Add and Remove events get sent out for the parent container.

[CxRes]

@joachimvh Can you point us to where this code resides in CSS?

[joachimvh]

@joachimvh Can you point us to where this code resides in CSS?

The DataAccessorBasedStore, which is the class that handles the "core" of data interactions as described in the Solid Protocol specification, is also the class that causes the initial trigger to happen for specific notifications, after which the MonitoringStore emits events that follow the path described here.

For example, this line causes the Remove event when deleting a resource from a container, and this line adds the metadata for the Delete event.

[elf-pavlik]

Create notifications get sent when a POST/PUT/PATCH request writes data to a resource that didn't exist yet. So the same situations where the required permissions would be different, as you require acl:Append permissions on a container when trying to create a new resource.

@joachimvh does CSS allow creating notification channels for topic which doesn't exist yet? #184

If it does not I think an Add notification would be sufficient if the parent container is the topic of the notification channel. Currently Solid doesn't support any other of adding a resource than creating a new one. Even if it eventually did, I don't know if a notification channel with a container as the topic has to communicate the (again currently non-existent) difference between already adding existing and newly created resource. The only situation where the Create notification would be required, seems to me if the notification channel is created (if we allow/support it) with at topic which doesn't exist yet.

[joachimvh]

@joachimvh does CSS allow creating notification channels for topic which doesn't exist yet?

CSS does no existence checks on a resource when trying to subscribe, so yes. The only check is if you have read permissions on the resource, so if the acl:default is set correctly on the parent container this is possible. Another possible scenario is when you subscribe to an existing resource, it gets deleted, and then it gets created again. The subscription will persist through those states so that last action will cause a Create event to be sent out.

The Add notifications are sent to clients who subscribed to the container, the Create notifications to the client that subscribed to the resource itself. So which one you get depends on what you subscribed to.

Note that CSS only sends Add when a resource is added to the container, not when one of its children is updated.

[elf-pavlik]

Thank you @joachimvh

I see at least two other issues relevant to what we discussed here:

• 184

• 185

You are very welcomed to provide feedback there as well.