jdev - 2023-05-12

If i have a groupchat with millions of messages and thousands of people. Can it be guaranteed by xmpp that each message can be referenced and fetched by a unique id that will always be unique? I remember the convo a few weeks back and it seems that there is such a thing. Just wanted to double check.

MSavoritias (fae,ve): messages in MAM are required to have a stable unique ID

typically UUID4 or similar is used by the implementations

https://xmpp.org/extensions/xep-0359.html

> MSavoritias (fae,ve): messages in MAM are required to have a stable unique ID I am interested outside of MAM ↺

My usecase is basically: A user was offline for a bit. And they can fetch all messages from other users in the groupchat based on that id

Fetching messages is done with MAM. I don't understand.

> https://xmpp.org/extensions/xep-0359.html Ah i see. And since the generating entity is the room server they are guaranteed to be unique then. Nice ↺

You should be able to use the stable/unique ID of the last message that you locally have to perform a MAM query to get all new messages 'since' that message.

I have no experience with how this performs on the scale that you're after. I expect that mileage will vary enormously, depending on implementation details.

And since every message has that unique id every message is unique in the context of the groupchat i am guessing

But yeah i will have to test it at scale

Yes, the (origin, stanza-id) combination should definitely be unique.

Nice.

Unless bugs or something, but that applies to everything.

> Fetching messages is done with MAM. I don't understand. I am interested if a mesh network kind of thing would work. Were you fetch messages from other users in a groupchat ↺

Hence my requirment of seeing if all messages are unique and can be referenced

(Independent of MAM)

If I was doing that, I'd still use the MAM protocol, I'd just query the other entities.

Well the uniqueness requirement is at the muc level

Having a single source of truth is nice, as I'm sure you either know or is about to find out :)

Probably problems with ordering yeah. Also need to know the authenticity and stuff. But still nice to experiment 😁

> If I was doing that, I'd still use the MAM protocol, I'd just query the other entities. Like each user? Hmm. Is it also used now that every user has a MAM database or is it only the server? ↺

But otherwise the idea sounds interesting

The MAM protocol could give me also the ordering so i would know what is missing. Especially if the ids ar unique. And checking with timestamps

Trusting a server component to not intentionally re-use IDs is one thing. Trusting thousands of users not to do that is a different thing.

True true ✏