jdev - 2024-08-23

  1. ubq323_

    rfc6120 seems to define <presence/> as a general 'publish-subscribe' mechanism, and <iq/> as a 'request-response' mechanism. so why do pep/pubsub (xep-0163/0060) use <iq> instead of <presence>?

  2. MattJ

    Because they have different semantics. I remember at one XMPP summit someone said "well, pretty much every protocol boils down to either pubsub or search"

  3. MattJ

    But the reality is that the details do matter, not every pubsub protocol has the same properties, depending on design decisions

  4. Zash

    Individual actions in PubSub, e.g. "publish", maps well enough to <iq>

  5. MattJ

    MQTT did not have any error handling until a very recent version

  6. MattJ

    <presence> does not have an acknowledgement

  7. MattJ

    <iq> does

  8. MattJ

    <presence> has additional semantics related to connectivity, which isn't what most pubsub applications want

  9. Zash

    So partially because most things also follow RFC 6121

  10. edhelas

    Somehow we should rewrite MUC on Pubsub one day

  11. MattJ

    Sounds like a great idea

  12. singpolyma

    Heh

  13. moparisthebest

    Plain XMPP is pubsub

  14. moparisthebest

    When you log in, you are subscribing to the "node" of your JID, then you can subscribe to other "nodes" by joining MUCs, you and others can publish to them

  15. moparisthebest

    As you can see, IRC is also PubSub

  16. moparisthebest

    https://burtrum.org/up/18f89623-7fd7-4cd2-b721-ef4d61ae8b35/91842w.jpg

  17. edhelas

    > New chat protocol just dropped > Looks inside > It's just pubsub

  18. edhelas

    https://upload.movim.eu/files/9d94237298995552fa13436420195fbca436dce7/ipO3M8Di3B56/image.png.jpg

  19. moparisthebest

    Accurate

  20. moparisthebest

    Redis: PubSub Kafka: PubSub Nostr: PubSub

  21. edhelas

    We need to do more XMMeMe

  22. moparisthebest

    Wait... That's not what we've been doing this whole time?