Thanks for the below. As a note, it's not really for Council to decide, it's for the consensus of the Standards SIG. So please, if you're reading this and have an opinion, do make it known. On Mon, 26 Oct 2026 at 12:25, Timothée Jaussoin <edhelas@movim.eu> wrote:
Hi,
So I had some discussions with Dave indeed to clarify some elements.
I'll list the concerns I understood and do some proposals for Council to decide.
The design that is proposed is this XEP is indeed focusing on simplicity and genericity. The idea is to embed the hierarchy information within the node itself. As defined in the introduction the goal is to make it reusable across various different "bookmarks systems" like:
Spaces node, like defined in 0503 PEP Native Bookmarks, defined in 0402 And hypothetically a future XEP that stores the Spaces subscriptions and where they can be reorganized and grouped together (as it is possible in Discord in the sidebar for example).
However this raises the concern that a node might have heterogeneous items in it. Here's what I found about it:
Pubsub (as 0060) doesn't forbid it, it can however be restricted using the 'pubsub#type' that I helped redefined a couple of years ago with pep with the following definition: "Specify the semantic type of payload data to be provided at this node." In XEP-0402 where it is not explicitly defined that the node MUST contains only '<conference xmlns='urn:xmpp:bookmarks:1'/>' items, however:
It can indeed conflict with the current specification regarding the items IDs that SHALL be the conferences JIDs (and here will be 'urn:xmpp:bookmarks-hierarchy:0' which can raise an error). Maybe some server issues especially with the synchronization with the legacy system And the parsing of unknown payloads while retrieving the node
So far I didn't found any clients that would behave badly if they see incoming 0402 items that are not '<conference>' items, but its just a wild guess and is indeed based on underspecification.
So my proposals for the Council regarding that issue:
Drop the genericity of Bookmarks Hierarchy and focus on Spaces only, we define this XEP as a way to reorganize Bookmarks items in a Space and maybe the future XEP that stores the Spaces subscriptions. XEP-0503 can be amended to mention that XEP and the fact that we can have different kind of payloads in the node items. Knowing that a Space node is stored on a generic Pubsub Service and the node id can be anything. Keep the current genericity but we would have to amend 0402 to define the fact that we could have different payloads in different items and see if it doesn't bother existing implementations. I think this is a more difficult path but it would allow clients to organize their Bookmarks in the future using the same spec (and will greatly simplify the implementation for clients like Movim that will handle 0503 and 0402).
Regarding Marvin feedbacks:
Regarding 2 and 4: This makes me wonder how to do UX when two ids map to the same title. In Roster Groups we don't have IDs, we just group by the title of the group, thereby entirely eliminating this issue.
Indeed we could add a small paragraph that says that its to the clients implementation freedom but that its advised to order it alphabetically on top, it can be even a SHOULD, then we keep UI consistency between the clients, but its a rare edge case.
Regarding 1 and 3: It struck me that there is two ordering mechanisms that are both independent and work differently (directories/groups ordered by an ordered list and bookmarks within a group ordered by assigned weights). It seems this can be collapsed to just one ordering, where you just order all bookmarks and then may decide to assign a group title to a subset of the bookmarks that is consecutive by weight to create a group.
Yes indeed, a first design had the same mechanism that you're proposing but it was creating some confusions and difficulties in the implementation, especially regarding the ordering of groups and setting their weights. The issues were:
We might end up with directories that have the same title but different weight due some synchronization issues When renaming a directory or changing its position we will have to update many items. From what I see in Discord you can have 20+ items per directory, even if its not that common the changes can create some spam on the network. But I understand that it will be kinda rare in the end.
However the pro is that it will only be <extensions/> based so totally avoid the concern above.
The third proposal for the Council:
Drop totally the '<directories xmlns='urn:xmpp:bookmarks-hierarchy:0'>' item in the XEP and indeed have a copy/pasted <directory title="Title" weight="3.2"/> in all the bookmarks items that need to be in the same directory.
It will be fully compatible with 0402 but will be a bit more different implementation in the clients, I'm not sure if its more complex or not.
--------------------------------------
So, to sum-up, the 3 proposals for the Council to decide:
Drop the Bookmarks Hierarchy for 0402 and only focus for Spaces, make it less generic but spec compliant Change 402 to allow different payloads including the '<directories xmlns='urn:xmpp:bookmarks-hierarchy:0'>' one Drop the '<directories xmlns='urn:xmpp:bookmarks-hierarchy:0'>' totally and instead add a '<directory xmlns='urn:xmpp:bookmarks-hierarchy:0' title="Title" weight="4.2"> inside each Bookmarks element to group them, making it generic for all the bookmarks.
I hope I've answered and clarified all the concerns and that the proposals fits with the need or limitations raised.
Thanks for your feedback and I'll adapt the XEP (and my implementation) based on what will be decided by the Council.
Regards,
edhelas
Le 26/10/2026 à 10:36, Dave Cridland a écrit :
On Mon, 26 Oct 2026 at 04:16, Nicolas Cedilnik <nicoco@nicoco.fr> wrote:
2) I'm a bit $thing about putting a hierarchy as a different item in the same pubsub node. What we've done previously is used a distinct node where the node id is derived from the node that "uses" it. So in the case of bookmarks, we might use urn:xmpp:bookmarks:1:hierarchy for example.
I was definitely expecting this feedback. With my gateway dev hat on, this does not change much for me. edhelas will likely chime in to defend this better than me, but I think it makes it more annoying for clients to make sure bookmarks and their categories stay consistent with a different node, and more IQs and pubsub#events to handle in general.
Edhelas sent me one or two messages last night, indeed.
Anyway, this is true, but I don't think existing clients expect to have heterogenous nodes - that is, a node where the items have different syntaxes.
They'll likely treat it as an error, and it feels possible they'd simply remove the erroneous item.
It's all, of course, a trade-off.
You could also inject the hierarchy directly into the bookmark extension data, giving its path, and letting clients figure out the hierarchy from there.
(Again, I wonder if a full bushy tree is the right option, as opposed to, say, tagging/groups).
Dave. _______________________________________________ Standards mailing list -- standards@xmpp.org To unsubscribe send an email to standards-leave@xmpp.org
_______________________________________________ Standards mailing list -- standards@xmpp.org To unsubscribe send an email to standards-leave@xmpp.org