Binding Feature - Preparing for non-Beta users.

[jvmahon]

I've been using the binding feature for several weeks now and belief this will be a great addition. However, at the moment, I think there may be several issues that suggest its not yet ready for a general rollout to less advanced users. Here are some key issues that may need to be addressed in moving from "beta" to production so as to reduce the amount of complex end-user education necessary for less knowledgeable users to use and manage the feature.

(I would, of course, love to hear I'm wrong about these issues and there is already a solution).

1. There are issues around the creation and deletion of Access Control List entries when entries are made via the new Bindings dialog box. These issues are being investigated, so hopefully they'll be solved soon. See Deleted Bindings also need to delete ACLs #1198
2. What happens when a node is deleted?
   • Assume you have 3 nodes. Node A has a binding cluster and is bound to nodes B and C. Nodes B and C have ACL entries to authorize control by node A.
   • If Node A is deleted from the web interface, it seems there needs to be methods that "clean up" any bindings or ACL entries in Nodes B and C. Specifically, there needs to be a function where, for each entry in Node A's binding cluster (i.e., the Node B and C entries), the code updates the ACL of Nodes B and C to remove the control entry for Node A.
   • If Node B is deleted, then the code needs to look at all the nodes bound to it (i.e., Node A), and update the binding cluster at Node A.
3. And What about node failure?
   • If Node X fails so there's no ability to read the ACL or Binding clusters of Node X, then there may be a need for a function that analyzes the ACL and Binding clusters of all the other nodes in the system to find any node (Nodes Y, Z) with a Binding cluster or ACL that references Node X and then update binding clusters and/or ACLs in Nodes Y, Z to remove the Node X references.
4. Startup check
   • Similar to 3, above, maybe there needs to be a function that, at python matter server startup, and/or on clicking a "check node bindings" button, the server does a check of all of the ACLs and Binding clusters of all nodes to make sure they are still valid and reference existing nodes.

Unfortunately, I don't know the Python Matter Server code well enough to try and implement this, but wanted to at least track the issues for future development (I also understand there will be a move to matter.js based server, so maybe these are issues addressed as part of that move).

[Apollon77]

Alright, here my opinions to that:

• Re 1.) Yes if you add ACLs on adding a binding you also need to remove the ACL entry when removing the binding :-) So this needs to be added for the removal case. Once the matter server got migrated to matter.js we will centralize this management (and also ACL optimizations and such) directly in the controller, so maybe better to wait some time.
• Re 2/3/4.) Exactly this I mean with "optimizing ACL" and such in my latest infos. For now that would all be needed to be implemented in the python side ... kind of react to node events (deletion mainly) and potentially cleanup bindings and ACLs and such

So with all that, and also given the fact that not that many devices with client clusters are out there I propose to wait till the matter.js based Matter server handles all this

[jvmahon]

Alright, here my opinions to that:

• Re 1.) Yes if you add ACLs on adding a binding you also need to remove the ACL entry when removing the binding :-) So this needs to be added for the removal case. Once the matter server got migrated to matter.js we will centralize this management (and also ACL optimizations and such) directly in the controller, so maybe better to wait some time.
• Re 2/3/4.) Exactly this I mean with "optimizing ACL" and such in my latest infos. For now that would all be needed to be implemented in the python side ... kind of react to node events (deletion mainly) and potentially cleanup bindings and ACLs and such

So with all that, and also given the fact that not that many devices with client clusters are out there I propose to wait till the matter.js based Matter server handles all this

Yes, I agree that this is something for matter.js.
I had originally posted this before the general release of Python Matter Server 8.1 as a warning that maybe 8.1 should not be generally released with bindings. But as it did get released with the function, it seems I was a bit late in this warning. Nonetheless, it would be great to deal with this in matter.js and I'm happy to help out then if I can. One thing to consider - the matter.js implementation will probably need to be able to "clean up" any problem caused by the Python Matter Server 8.1, but I believe the functions 2/3/4 I outlined should do the trick.

The only devices I'm aware of with binding clusters (so far) are the Inovelli series of switches. I've occasionally checked the compliance documents for product on the www.csa-iot.org and haven't noticed others with the binding clusters.

[marcelveldt]

Considering the fact that this binding feature is only offered as experimental feature in the debug UI only, I wouldnt worry too much about this. 99% of the users wont even know this feature exists.

Long term goal is to make bindings robust and user friendly and offer a direct way to configure them from HA interface.
For now, the current implementation is a great way to experiment

[noranraskin]

@jvmahon How do you find devices with the binding cluster? Like the inovelli switches (which to my knowledge are not sold in europe). Would like to get one for testing

[jvmahon]

@jvmahon How do you find devices with the binding cluster? Like the inovelli switches (which to my knowledge are not sold in europe). Would like to get one for testing

Few products seem to allow binding as a controller. I don't think there is a easy way of figuring out which ones do and which don't. The only way that I could think of would be to download the products Compliance Document (e.g., from the CSA web site, as an example: https://csa-iot.org/csa_product/smart-on-off-switch/) and check the endpoints for ones with Binding Cluster test plans.

[cedricziel]

Hi, i created a custom integration to manage bindings and their respective ACLs, because i found it too cumbersome with the current UX of the matter server. - The component handles provisioning bindings and appropriate ACLs.

I'm contemplating contributing the UX to this component, but given its in maintenance mode i worry the effort is lost - and the new implementation on top of matter.js is not public yet, is it?

https://github.com/cedricziel/ha-matter-binding-helper

[Apollon77]

@cedricziel In fact the new version should become public very soon. and in fact we just take the current dashbord over "as is", so it should be easy to PR the changes then. give me some more days

[cedricziel]

Very nice - let me know how I can help or when you want someone to test-drive the new server!

[Luxni]

@cedricziel
Why not integrate directly into HA's matter component?
I think it would be better to display this information on the matter device page.

[cedricziel]

Hm. I have two thoughts:

1. I do not like the overly generic device page in HomeAssistant - as a user, it's just a giant blob of stuff that's not overly well organized (and I acknowledge that this seems to be by design)
2. doing it in the matter server gives all users, HA users and non-HA users the option to bind easily

Taking the idea of the core integration a step further: I would also love to manage matter Groups and Scenes - currently matter server doesn't provide enough API for a clean approach but as soon as it does it might be a good idea to brainstorm entity design - like how would matter scenes and groups work with HA and vice versa. A part of this should stay truely uniform for sure as is most of HA, but matter poses an opportunity to be more opinionated for some of this (as ZigBee might as well).

My goal for the integration was to use bindings to become more independent from core HA downtime and network turbulences and to implement it so it would "just work" in matter.