Proposed XMPP Extension: Portable Import/Export Format for XMPP Components
The XMPP Extensions Editor has received a proposal for a new XEP. Title: Portable Import/Export Format for XMPP Components Abstract: This document specifies a file format for importing and exporting component data to and from XMPP software URL: https://xmpp.org/extensions/inbox/pie-components.html The Council will decide in the next two weeks whether to accept this proposal as an official XEP.
On Sun, 25 Oct 2026 at 05:44, Daniel Gultsch <daniel@gultsch.de> wrote:
The XMPP Extensions Editor has received a proposal for a new XEP.
Title: Portable Import/Export Format for XMPP Components Abstract: This document specifies a file format for importing and exporting component data to and from XMPP software
This has little of any concrete nature, and according to my current understanding of Council rules applied to ProtoXEPs should therefore be rejected. First, as described, this is useless for both pubsub and MUC. It goes as far as describing an "item" subelement, but then fails to explain what this represents for either. One assumes it might be a room for MUC, and a node for pubsub, but what would then go inside each is entirely unspecified. It feels like, for any kind of useful schema to be produced, that having explicit "room" and "node" elements would be more useful (in which case this specification is essentially useless). Alternately, mandating a single payload element, or specifying at least that only the first was the payload element, seems essential. Second, even if it had covered pubsub and MUC at all, it fails to cover the "other services" it claims to. This is, to quote, simply a "document that says things". This is a shame, as solving the missing MUC and Pubsub support in PIE would be a useful thing for the XSF to work on and would indeed fill an important gap. If it were to be accepted counter to these previously stated rules, then I'd recommend the following while experimental: 1) Drop the "name" of the item, and replace it with "jid", "node", or both, to match disco#items. I would be inclined to specify that the first child is the payload (mandatory to understand), and any others are optional extensions. 2) Drop the type of the component in favour of "category" and "type", again from XEP-0030. 3) Get at least the basics of MUC and/or Pubsub. I'd think MUC is config and affiliations, and Pubsub is config, affiliations, subscriptions, and items. I think all of those can be in the form they'd be returned by relevant IQ. Seems easy enough as a first cut, though undoubtedly implementation and discussion will surface more. All these get wrapped in a payload element to act as a schema support. Dave.
First, as described, this is useless for both pubsub and MUC. It goes as far as describing an "item" subelement, but then fails to explain what this represents for either.
I agree the MUC example for item could be made more clear that the name in this case is the JID localpart. Or perhaps instead of saying "type specific" we could mandate that it is always a JID localpart. That would be fine for all my current implementations, but I wanted to leave room open for other types in the future to want name to maybe be something else. But honestly I'm not sure what else it could be so maybe saying it is always a localpart is fine.
One assumes it might be a room for MUC, and a node for pubsub, but what would then go inside each is entirely unspecified.
For MUC it gives a rather specific example of a <archive xmlns="urn:xmpp:pie:0#mam"/> child. Pubsub wouldn't use item tag at all, as it says it would use <pubsub xmlns="http://jabber.org/protocol/pubsub#owner"/> child just like existing PIE does for PEP.
1) Drop the "name" of the item, and replace it with "jid", "node", or both, to match disco#items. I would be inclined to specify that the first child is the payload (mandatory to understand), and any others are optional extensions.
I think calling it "jid" is confusing since it's just the localpart, and also component already has "jid" attribute. But I do think we can be more specific about what "name" means, per above.
2) Drop the type of the component in favour of "category" and "type", again from XEP-0030.
This is an interesting idea, and I think I quite like it. Though I'm concered that we don't actually have an identity category/type pair that uniquely identifies MUC (we have conference/text for example, but that could apply to any conference protocol).
3) Get at least the basics of MUC and/or Pubsub. I'd think MUC is config and affiliations, and Pubsub is config, affiliations, subscriptions, and items. I think all of those can be in the form they'd be returned by relevant IQ. Seems easy enough as a first cut, though undoubtedly implementation and discussion will surface more. All these get wrapped in a payload element to act as a schema support.
My current implementations hvae just archive and items, which is why that's all I specified in the document, but having eg the MUC config data form as another optional child in that case, etc, as you describe seems very sensible to me.
On Sun, 25 Oct 2026 at 19:36, Stephen Paul Weber <singpolyma@singpolyma.net> wrote:
First, as described, this is useless for both pubsub and MUC. It goes as far as describing an "item" subelement, but then fails to explain what this represents for either.
I agree the MUC example for item could be made more clear that the name in this case is the JID localpart. Or perhaps instead of saying "type specific" we could mandate that it is always a JID localpart. That would be fine for all my current implementations, but I wanted to leave room open for other types in the future to want name to maybe be something else. But honestly I'm not sure what else it could be so maybe saying it is always a localpart is fine.
I mean, it's entirely unspecified in this, for MUC or any other case.
One assumes it might be a room for MUC, and a node for pubsub, but what would then go inside each is entirely unspecified.
For MUC it gives a rather specific example of a <archive xmlns="urn:xmpp:pie:0#mam"/> child.
Yes, but this is explicitly, as you note, an example. Lower-case "may", and all. Again, there's no actual specification here.
Pubsub wouldn't use item tag at all, as it says it would use <pubsub xmlns="http://jabber.org/protocol/pubsub#owner"/> child just like existing PIE does for PEP.
No, it doesn't say that, it says it might. Lower case may, explicit example, etc.
1) Drop the "name" of the item, and replace it with "jid", "node", or both, to match disco#items. I would be inclined to specify that the first child is the payload (mandatory to understand), and any others are optional extensions.
I think calling it "jid" is confusing since it's just the localpart, and also component already has "jid" attribute. But I do think we can be more specific about what "name" means, per above.
Well, obviously I'd change it from being just a local-part (or type specific value, as it is now) to being explicitly a jid as per disco#info. Yes, this absolutely duplicates information, but it's doing so the same way that disco itself does. This being an email, not a ProtoXEP, I feel I'm allowed to hand-wave a little, but my thinking is that if we do this, then we can also place the item element at the server level of the PIE for various odd-ball cases.
2) Drop the type of the component in favour of "category" and "type", again from XEP-0030.
This is an interesting idea, and I think I quite like it. Though I'm concered that we don't actually have an identity category/type pair that uniquely identifies MUC (we have conference/text for example, but that could apply to any conference protocol).
In principle yes, but in practice no. MIX, for example, uses a different type.
3) Get at least the basics of MUC and/or Pubsub. I'd think MUC is config and affiliations, and Pubsub is config, affiliations, subscriptions, and items. I think all of those can be in the form they'd be returned by relevant IQ. Seems easy enough as a first cut, though undoubtedly implementation and discussion will surface more. All these get wrapped in a payload element to act as a schema support.
My current implementations hvae just archive and items, which is why that's all I specified in the document, but having eg the MUC config data form as another optional child in that case, etc, as you describe seems very sensible to me.
Again, you didn't specify, you merely suggested that a specification might suggest that. Now, in my view, this is a fine start to a useful XEP, and it's worth getting it into Experimental to work on. But in its current form, it does not specify very much at all, and cannot be concretely implemented. It's a shame that current members of Council are so vocally against such things. Dave.
participants (3)
-
Daniel Gultsch -
Dave Cridland -
Stephen Paul Weber