2020 in

Topic 1: Shortage Audit

We talk about ways in which xmpp comes short compared to other solutions like Slack, WhatsApp etc.

On the list are things such as contact discovery, snippets, microblogging ("stories"), stickers, bots and other integrations.

Some of those are already specified but lack implementation, others are not yet tackled at all.

Topic 2: Easy Password-less Onboarding and Account Management

User onboarding could be vastly simplified by having the registering device generate a random password/key. Problems arising from this are usage with multi devices (QR code scanning with existing devices), password reset (reset link via email, pin via other device).

Other things to consider are client certificates (XEP-0257) and other SASL authentication methods.

Topic 4: Rich Account Presence in PEP

Classic presence in is device centered and breaks down in multi device usage.
Eg. if the user sets his desktop client to "away" and starts using his phone, is he now away or on the phone? Both?

Publishing the status of a user (eg. "I'm at the Summit \o/") to a PEP node would solve this issue as that node would be per account instead of per device. The latest status would override the previous one.

This would also allow for rich, microblogging like status sharing as known from WhatsApp or Instagram "stories" although status and stories would probably be different features.

Using microblogging (XEP-0277) would be one possible solution for the stories thingy.

Topic 5: Message Layer Security

A workgroup within the is currently developing a new end-to-end encryption protocol called MLS which is basically the IETFs answer to the Signal protocol.

MLS focuses on group chat encryption in larger groups, but is also suitable for normal chats. Dave gave an overview of the protocol and suggests that we start looking into how to deploy it with XMPP now.



Topic 6: Easier End-to-End Encryption Key Management

Key Management in e2ee systems like OMEMO can be a pain, especially in group chat scenarios.

Proposed solutions are cross signing between keys, identity key sharing and/or master identity keys.

The discussion participants tend to agree that the master key approach is an elegant solution.

Quick coffee break. Syndace's recommendation of the day: Mix coffee and chocolate in the same cup!

Lightning talk of Alex about his cross platform XMPP client Palaver.
The app is developed using the MVVM pattern and runs on Windows. Definitely worth to keep an eye on!

Topic 7: Why Push Notifications are not enough?

Push on iOS sucks and appears to be unreliably. Additionally the Push XEP-0357 lacks support for different types of priorities. Therefore spam messages from strangers will always alert the user as there are no silent notifications.

Tigase proposes some very interesting solutions to current push issues, like using encryption to hide message contexts inside push notifications from the push service itself.

There needs to be server-side tweaking in order to figure out which stanzas should trigger a push notification and of what priority.

There is a flowchart diagram by the slack developers that gives an insight on the complexity of push related decisions.

Criteria that need to be taken into consideration are:
* mentions / direct messages
* calls
* notification preferences per channel
* e2ee
* device type

Topic 8: MIX, when?

MIX is the spiritual successor of MUC group chat. There exist some implementations that suggest that implementing it is feasible. For wider adoption however, MUC downwards compatibility would be necessary.

Part of the feedback to the specification has not yet been addressed, eg. where to store records of all the MIX channels a user joined (roster?).

There is also the idea floating around to add a the id of the previous message to each message, so that S2S sync issues can be detected and fixed.

Some GDPR issues with message sync are being discussed, but the discussion has to come to an end for today.

This concludes the first day of summit :)

Topic 1: IM NG

Historically, XMPP messages were routed to only one receiving device. That may be the device with the highest priority, or a particular device identified by the resourcepart of the jid (eg. vanitasvitae@jabberhead.tk/phone is a phone).

To improve the user experience in modern multi client scenarios, extension protocols like Carbon Copies and Message Archive Management were introduced to adjust the message routing to changed user expectations.

However, this still leaves some uncertainties in how messages are supposed to be routed in certain situations.

XEP-0409 IM Routing NG aims at solving this issue by simplifying and unifying message routing for enabled clients. Basically messages sent to a barejid are routed to all IM NG devices, while fulljids are no longer being used in the instant messaging context (you can still use fulljids for other scenarios like machine to machine communication of course).

Currently there are some edge cases to be discussed like routing of error messages. Those are being discussed right now.

One proposed idea is to simplify the rules of Carbon Copies until Carbon Copies "transforms" into IM NG (which has a simpler syntax). That way the complex rules of Carbons can be dropped some day.

Topic 2: Lightning Talk of @nyconyco about the work of the Comm Team.

Comm Team is doing the monthly XMPP Newsletter (to which you can subscribe to by the way: xmpp.org/newsletter.html )

Show newer

@vanitasvitae Could you briefly give some examples of what the issues are with Carbon Copies?

@stevenroose Getting it to work reliably with MUC is apparently non-trivial. Also on login there is the possibility of race conditions between carbon enabling and other setup steps.

@vanitasvitae @tigase Hmm, that's not entirely true. The user shouldn't be notified when a push notification comes in. The app should be woken up and the app should check what's new and then decide whether the new content is worth waking up the user, right?
That content doesn't have to be part of the push necessarily for that.

@vanitasvitae What's the difference between identity key sharing and a master key? If the ID key is shared, isn't it a master key?
(I know I could go try read the minutes, but it might be helpful to have here on the record as well.)

@stevenroose a master key would be eg. an OpenPGP key which signs all OMEMO keys. The benefit of this is that the master key can be used to sign the identity keys of different encryption mechanisms.

@stevenroose It doesn't have to be an OpenPGP key though. If it is some sort of certificate, it could also be used to log into the account in the first place.

@vanitasvitae Oh that'd be cool! Logging in with a PGP key by signing a challenge!

Sign in to participate in the conversation

Fosstodon is an English speaking Mastodon instance that is open to anyone who is interested in technology; particularly free & open source software.