On 7 October 2025 22:23:27 CEST, Peter Saint-Andre <[email protected]> wrote:
>On 10/7/25 1:52 PM, [email protected] wrote:
>> thanks Peter. interplanetary means: long delays and disruptions. Earth to
>> Mars one way delay = 4-22 minutes. So a messaging/protocol should not be
>> "interactive" by having all kind of updates which are not real messages.
>> Moreover, the transport has to be compatible. TCP won't work with this use
>> case, so we are using a space profiled QUIC. So we would need XMPP over
>> QUIC, which AFAIK is not active.
>
>Hi Marc, thanks for the additional context.
>
>It's possible that some folks might be interested in XMPP over QUIC for other
>reasons, so that might be a good place to start if those folks (whoever they
>are!) have the energy to work on a spec for it.
Indeed. When I was working at VEON, had our project continued, I'd added
support for XMPP over QUIC in our mobile apps. The primary reasons where:
built-in TLS, 0 to few roundtrips for reconnection, and switching between
networks while retaining a (logical) connections. Oh, and multiplexing and
co-streaming with RTP, STUN, TURN. So, both for XMPP as well as voice calls.
For the latter, consider the following. I was working in an office building for
which "our" wifi was on the 5th, 4th, and ground floor. When going for lunch
and having a call, I would have WiFi on the 5th, step in the elevator, lose
wifi, so switch to cellular. Elevator stops on the 4th, too, briefly switch
back to WiFi, doors close again, cellular, ground floor WiFi, exit the building
and end on cellular again.
With QUIC you can retain the connection across those roaming changes, with
virtually no disruption.
I'd love being (professionally) involved in such a project again.
I totally see the value defining this at the IETF, and the interplanetary angle
makes a lof of sense to me. For regular messages, XMPP has great semantics in
its protocol because it doesn't require acks for the most basic stuff.
Obviously some features will, as of course iq stanzas.
--
ralphm
_______________________________________________
Standards mailing list -- [email protected]
To unsubscribe send an email to [email protected]
