XSF Discussion - 2019-01-03

s2s:show('hardteckno.com') | OK: Total: 60 outgoing, 48 incoming connections

bug? feature?

it's the exact same numbers as if I did without the domain, just that the connections don't get listed

oops.

wrong room

Why is "Simple IoT Client" listed in the XMPP Clients list, again?

It also looks like its link is broken.

broken link -> expire it immediately

What can I do to expire Pidgin?

> waher.se took too long to respond. Might be a temporary failure.

> broken link -> expire it immediately how long do I need to DDoS pidgin.im to get it removed?

hrhr

Ge0rG: Try it and let us know how long it takes.

is `<span style=" font-weight:600;">` correct XHTML-IM for bold?

If you allow style

how is a client supposed to know that 600 = bold?

that’s how bold is defined

bold is just an alias for 600 or something

Ge0rG: you need a `</span>` for it to be valid

bold Bold font weight. Same as 700.

https://developer.mozilla.org/en-US/docs/Web/CSS/font-weight

So 600 is not-quite-bold?

600 Semi Bold (Demi Bold)

poezio will display as bold if you have font-weight:anything in the CSS

font-weight: normal == 400

m(

Check out values here: https://developer.mozilla.org/en-US/docs/Web/CSS/font-weight#Values

https://developer.mozilla.org/en-US/docs/Web/CSS/font-weight#Common_weight_name_mapping rather this table, no?

Yeah

That's a nice piece of documentation

Now I remember again why I hate HTML

Ge0rG: Why exactly? :)

first, this is CSS

second, what’s wrong with it? ✏

jonas’: CSS is a part of HTML.

You know what they said about PHP? A fractal of bad design.

CSS is commonly used with HTML, but you can use HTML without CSS just fine, and you can use CSS with things which are not HTML (e.g. GTK or SVG) ✏

I don’t see that here though

Ge0rG: You need to make peace with the fact that everything sucks, and that is unlikely to ever change :)

waqas: I can't make peace with it, I can merely try to rant less.

who’s responsible for the registries? (<https://github.com/xsf/registrar>)

jonas’ Until there is a perceived need for a more formal governing body, the functions of the XMPP Registrar shall be managed by the XMPP Extensions Editor [6]

https://xmpp.org/extensions/xep-0053.html

XSF Board Meeting | Logs: http://logs.xmpp.org/xsf/ | Agenda https://trello.com/b/Dn6IQOu0/board-meetings

0. Welcome + Agenda

Hi!

nyco sent regrets

Guus, thx :)

Hi

hello

MattJ?

Anything to add to the agenda?

Not me

I just added things to Trello

trademark, email server status

Ok

Me too

1. Commitments

Vacation is almost over here, making more time for all things XMPP this month.

Including finally getting the items with Peter sorted.

2. FOSDEM / Summit

Guus, any news on hotel?

I've send a request for a quote, but have not received one yet.

I expect that to happen today or tomorrow

after which I'll forward it to the mailing lists, much like we did last year.

Otherwise, let's sync tomorrow on all the things

(I'm getting a quote from Thon EU again)

I've also tried to reach out to the same restaurant for the XSF Dinner

couldn't get someone on the phone, but left a message

that's it for now.

Ok

Thanks

3. GSoC

Thank you Guus

Joachim expressed some interest in participating, but communication seems to have broken down over the holidays.

(GSoC, that is)

I'll follow up with him

no others have stepped forward.

Let's aim to have a go/no go in next weeks meeting?

Ok, maybe good to repeat the request now holidays are over

I don't like battering people. I'll publicly follow up Joachim. If someone else is interested, they can chime in.

Ok

Hey

Good

Sorry, here now

5. JabberSpam trademark

hi MattJ

(hi)

Good comments, Guus

Ge0rG has send in an application, that has had little response. He requests action.

I'd still like to get guidance from Peter

Peter acknowledged my request some two weeks ago.

Yes, I got a copy

interestingly, the website speaks of a Trademark WT

who's that?

IIRC, last time I asked for a trademark license, it ended up being voted by Board (after Peter's principal approval)

Currently, just Peter, I think.

(It does not explicilty name it a work team, but it suggests that there's a group of people, plus the executive directory, that are said team).

Director

sorry 🙂

The agreement does mention a "trademark committee" iirc

that might be it, yes.

Adding it to the list of topics.

Ge0rG: trying to get that resolved soon

I just created a small PR to the website, that should get Peters attention too

(regarding pending trademark applications)

Hopefully, we can gain some traction that way too.

6. E-mail issue for seve

I'm not sure if this is just for Seve

Further discussion has shown that I might need _two_ trademark permissions actually, one for the Org (requested), and another one for the "Jabber Spam Fighting Manifesto"

I saw some discussion and request to remove from RBL

I don't know what RBL is - or if we indeed do have an issue

Seve: did you get nyco's email?

but for several weeks, people seem to have email related issues

Seve is one, but mail from the wiki (on account creation) do not show up either

unsure if it is related

If this keeps up we may have to start sending through a service like MailGun, I'll ask the iteam what their strategy is.

ralphm: still no new emails from XSF lists, I was thinking on waiting for a new email to check if I get them now

I'm hoping that iteam can give some kind of status update.

Seve: ok, that was sent just before this meeting

if only to confirm or reject the notion that we have issues.

Kev, intosi?

ralphm: then no, I still do not get them

I think someone will have to check the mail server log again then

Aye

Ok, taking that up with iteam.

Thank you for this, I really appreciate that

7. AOB?

I have one AOB

no AOB from me.

None here

Tomorrow is our 20th birthday. Somebody should give a party. https://slashdot.org/story/99/01/04/1621211/open-real-time-messaging-system

Indeed.

Of course the party will be distributed, with Disco and lots of Jingle.

:)

ralphm: are you going to MIX the drinks?

musthinkofaMIXjoke...

thanks.

Ge0rG: sure. I'm more Pub than Sub.

that sounds rather zimpy.

any practical idea's on commemorating the milestone?

apart from bad puns, obviously.

Guus: somebody should write a blog post. I suggest "the half-life of instant messengers"

I had great ideas and no time, so that didn't work out.

I think we wanted to organise one with Nÿco this year.

I'd volunteer, except -EBUSY

that goes for everyone, I'm afraid.

But we might be able to do something around the Summit

maybe we can crowdsource it? Collect the lifespans of IMs in a pad

Where does gajim keep the passwords on Windows?

I can manage an hour or two tomorrow to write it down

Oops

Sorry, wrong chat

lol https://slashdot.org/comments.pl?sid=15607&cid=2048739

Hehe

Ok, with that.

8. Date of Next

+1W

wfm

9. Close Thanks all!

+1

until we meet again!

Thank you!

XSF Discussion | Logs: http://logs.xmpp.org/xsf/ | Agenda https://trello.com/b/Dn6IQOu0/board-meetings

Zash: XMPP, a story of NIH

Ge0rG: All of humanity probably

So does anyone volunteer to collect data about IM networks/apps and their lifetimes?

https://en.wikipedia.org/wiki/Instant_messaging#History

Zash: that's very coarse

but maybe a full history of all abandoned networks will be less funny of a read than I imagine

Oh, https://waher.se/IoTGateway/SimpleIoTClient.md is back up

ha I didn't know that "The term "Instant Messenger" is a service mark of Time Warner[11] and may not be used in software not affiliated with AOL in the United States."

wtf is that true??

Things like that are why we ended up with the term "roster", when at the time everyone was talking about your "buddy list(TM)" (e.g. https://www.bizjournals.com/sanjose/stories/1999/05/31/story7.html )

Also why we ended up with XMPP.

Trademarks are why we can't have nice things

trademarks don't expire, right?

They do

No they don't

i.e. if you register a trademark you have to renew it after ~10y

Right

Which they'll do, forever

http://tmsearch.uspto.gov/bin/showfield?f=doc&state=4802:35rtkj.4.26

> This search session has expired. Please start a search session again by clicking on the TRADEMARK icon, if you wish to continue.

It's just the "BUDDY LIST" result, it's still registered to AOL

You also have to actively protect it as well, right? Ie go after people using it without permission and stuff.

Hm, but then I'm not sure which is whic hof ™ and ®

🤷

https://slashdot.org/comments.pl?sid=15607&cid=2048734 "clients are quite easy to write", fast forward 20 years later

Clients are indeed easy to write. It's just good clients that aren't.

true

also 20 years ago there was no MAM and Carbons no phones etc

no encryption, so it was basically, download the roster, and send a message

Hi, happy new year everybody. In XEP-0060, if I have an item with id "abc", I publish an other item with it "def", then I publish a new item with the first id ("abc") which will overwrite it. if I then request items with max=1, should I get "abc" or "def" ? § 7.1.2 says that item is overwritten and § 6.5.7 says that items returned are the "most recent". So I guess it should be "abc", right ?

ralphm: ^

I think that question was also raised by edhelas a few months ago(?) I don't know if there's a clear answer

If you think about it as publishing a new item that just happens to also delete an older item, then it makes sense that the 'abc' one is the last item you get

I'd argue, without looking at the xep, that something that's overwritten is not 'new'

I got the same 2 thoughts, so it's confusing because 2 options could make sense.

the XEPs states that the mosts recents items must be returned, so even if you overwritte, the "abc" one is the more recent.

most*

The identity is not new

yes, but the item is

Is it new, or is the old one changed?

I prefer the way where I don't have to throw out all the append-only assumptions from everywhere

So I've coined this idea a few times the last few weeks in random MUCs, but I'm not sure how to approach taking it further than an idea: I (and I'm sure others) have been thinking quite a bit about OMEMO key fetching and how easy it is for server admins to just serve extra keys for contacts etc. I don't think there is a single client that does not automatically accept all keys by default. (Conversations has an "expert setting" that lets you turn of accepting new keys. I think Gajim has something similar.) I've been thinking about PGP to help improve this. My personal main objection to using PGP for encrypted messaging is that I prefer to not have my private key on my device at all times (in unencrypted form) like you need for XEP-0374. Instead, one could sign OMEMO keys with a PGP key to just have to do this once for each new device. In theory, this would not need to have your PGP key on a mobile device, for example. Since you could verify the OMEMO key fingerprint on on your desktop and then sign it there. On the mobile device you only need to import your own public key and signed public keys of your contacts.

Hah, Syndace ^

Not sure I'm missing something that makes this hard to use. Also I don't know if PGP is still used at all.

why not encrypt the messages with pgp?

We've been discussing with Syndace a bit and trying to find solutions about your concerns on the server being able to inject devices etc.

oli, because this needs the pgp private key to be available at all times

OMEMO keys are single-use-case and can easily be replaced when confiscated

The idea with PGP is that the key would be stored on the server and the client can unlock it, but that has other pitfalls

A PGP key is kinda like your ultimate beacon of trust 😀 We use it a lot at work f.e. for automatic deployments etc

So I never have my laptop or phone have it unencrypted and need to enter a lenghty passphrase for every use.

(Well technically it could be done any way, but that's what I hear the most, that makes the most sense UX-wise)

I don't think it's nice to type a passphrase for every message 😀

Not for every message

steven: good idea, but this would require OpenKeychain on Andoird to verify the signature and/or sign the statement

pep., I don't know how XEP-0374 works, tbh. Does it just use one master key all the time? Or does it use ephemeral subkeys or so?

Wiktor, to verify yes. But to sign your own mobile key, you could do manual fingerprint verification with a desktop client like Gajim and sign your mobile's OMEMO key there and send the signature to the server. (Just thinking out loud here, though.)

You choose? I don't know it that much either, I'm definitely not the reference here. I also know other people have concerns about 374, but I'm waiting on them to tell because I don't have the knowledge to back these claims

Yeah, actually Conversations already has similar code but using X.509 instead of OpenPGP

steven: so you want cross-signing basically right

I think the way you're trying to implement it is going a bit far

pep., yeah well it's also possible of course to sign on the mobile client

still you'd have to enter the passphrase only once

instead of very often/every message?

I saw you proposing that before but I didn't see a way to do that in a way which is not overkill.

But now that I think about it again you could probably do it without too much complexity

You might not even need GPG itself, rather a master key of any soet

But I'm busy right now, I'll take some time to think about it later/tomorrow

Syndace, well, "a master key of any sort" isn't much better. The thing is that quite some people already have some form of web of trust with PGP keys and verified identities. (The company I work for is fully remote so at our annual offsite we do a quick PGP key signing ritual. From then on we can f.e. introduce a new coworker by having him meet a single colleague that signs his key.)

Basically PGP is identity-based while OMEMO is device-based. So to tie a device to an identity, it makes sense to use PGP I think.

steven: PGP is a can of worms, especially but not exclusively regarding UX. Not even hardcore cryptowhores figure out all of its quirks

I like the matrix idea of a master olm(?) key.

Ge0rG, true. But it's an accepted default.

Ge0rG, many people say the same about XMPP 😀

No need to mix different crypto libraries with each other.

steven [19:58]: > Ge0rG, true. But it's an accepted default. Nope. S/MIME is the accepted default.

The PGP web of trust is just silly. I've verified your identity, therefore I trust you to verify other people's identities?

I think that PGP has a place in xmpp indeed, but without OMEMO then.

Just have an account key, exchange it with your friends, share it between all your devices, problem solved. You leak your key? All of your chat history is compromised.

You lose your device? Lucky you if you still have the key / recovery password. Then you'll regain all your logs.

OMEMO trust management is just madness. What do you do if you verified one of your friend's devices, but none of your own other device keys?

It barely works as long as you have exactly one device and it doesn't get lost, stolen or broken.

Ge0rG, I don't think you have much experience using OMEMO..

I have the Conversations "paranoid mode" where I have to manually approve new device keys and it works fine.

I don't like the whole idea of omemo/otr. The only improvement in it over gpg is PFS but too many drawbacks. And gpg is good enough to stop any realistic state wide spying efforts. So PFS is needed to those who REALLY has reasons not to be spied and MitMed and traffic decrypted, and we know all too well who these people are. :-/

When I first start chatting with a new contact, I will just blindly hit "ok" (I'm not gonna call them to spell it out for me), but after that when I get sent new device keys, I just ask them first if they started using another client.

So yeah in theory the admin could still hijack the key on the moment someone starts using a new client. That's why I'd prefer to just have my contacts' PGP keys and have them sign their OMEMO keys.

So, which keys could admin hijack?

steven [20:05]: > I have the Conversations "paranoid mode" > When I first start chatting with a new contact, I will just blindly hit "ok" (I'm not gonna call them to spell it out for me) I rest my case.

If he hijacks your public keys, then what?

Andrew Nenakhov, the admin could install a module that whenever a user adds a new device, it broadcasts a different key instead that it owns itself. Because I described that I would only ask "did you start using a new client?" without also verifying the fingerprint.

Ideally I just send them the fingerprint using their first OMEMO key to verify.

Andrew Nenakhov: the server Admin could add another device key to your account, or replace your key with his own.

Andrew Nenakhov, he could but only if he's already doing that at the moment of the first encounter.

steven: how do you ask your friends whether they got a new device? With the old key? Via SMS?

Ge0rG, with the old key(s).

Usually it's someone that opened the webchat for the first time or downloads a desktop client or so.

steven: so if they lost their phone, you are out of luck.

So yeah I should ask them to verify the fingerprint. But I don't have such highly sensitive conversations yet. Just thinking that in case I have, I'd prefer PGP instead of manually messing with fingerprints.

Ge0rG, if they lost their phone and have never used a desktop/web client, yes.

how do you verify their PGP key though?

steven [20:11]: > in case I have, I'd prefer PGP instead of manually messing with fingerprints. Now with *that* I can totally agree.

(Also note that I'm the server admin of the server my social network is on, so I should have been targeted by a hacker for shady things to happen.)

moparisthebest, well, you only have to do that once. And you could delegate that to people you trust to do it thoroughly.

Also for higher-profile people, their PGP keys might be publicly known and signed by a bunch of people.

steven, that what fingerprints check is for, so you should verify your contact fingerprints via an independent means of communication.

You already specify your own PGP key in C, one can check if your contacts PGP key is signed by you

Andrew Nenakhov, or with a signature of an authority you trust.

Cool. So this authority could be compromised and all your struggle and pain with encryption will be for nothing.

There is no trusted authority on PGP. This is what S/MIME is for...

Like say some guy from The Guardian contacts you. He uses an OMEMO key. Most likely, his PGP key will be known, online on several websites and signed by people from other newspapers etc. If he signs the OMEMO key with that PGP key that I can find in multiple places with multiple signatures from other keys I can find in even more independent places, I would personally rest assured.

It never ceases to amaze me how people want security and privacy but not the inconveniences that mandatory come with them.

Andrew Nenakhov, there's several levels of privacy of course. Of course I'd like the conversations with my friends to be private from petty hackers and bad admins getting government orders. But I know that these conversations are not safe from high-profile cyberspecialists. That's fine. If I'm about to become a whistleblower and talking with a newspaper, I'll up my security and me tolerace to the nuisances that come with it.

> Ge0rG> There is no trusted authority on PGP. This is what S/MIME is for... Trusting that authority is another story. DANE anybody? Does S/MIME even work with that

steven: you've heard of https://evil32.com/ already?

pep.: there was a proposal

I'd love to have an implementation of that.

pep.: but not just the fingerprint, store the whole certificate in DNS

> steven: you've heard of https://evil32.com/ already? Ge0rG, hmm, I don't use the shortIDs personally. Not sure how, but my `gpg --list-keys` prints full IDs.

steven: the point is that the key of your journalist is fake, together with all the keys that signed it

steven: defaults of gpg change over time, no automated system should use short fingerprints (OpenKeychain follows this)

Ge0rG: not necessarily, first of all legacy sigs used long key ids not short 32 bit but for years the full fingerprint is embedded in the signature

Why isn't anyone complaining that HTTP upload to a MUC exposes your domain to all muc participants?

Ge0rG, because Conversations displays a picture instead of an URL.

Wiktor: Chance fifty fifty

your avatar exposes things too

So people are not aware of that.

probably a bunch of other things

moparisthebest, uh, no, it doesn’t.

in a different way, it lets me tell 'dwd' in one channel is the same as 'Dave' in another channel etc etc

if I happen to have the same person in my roster, that too

Everybody should use the same avatar!

Ge0rG: this is 4 years old: https://gnupg-devel.gnupg.narkive.com/Z0EFUBU7/issuer-fingerprint-was-vanity-keys

Wiktor: I'm speaking about obtaining a key out of band

> Wiktor: Chance fifty fifty > Wiktor: I'm speaking about obtaining a key out of band ?

OpenKeychain uses qr codes, full fingerprint

But you can't scan the fingerprint of some journalist

This one uses full fingerprint https://theintercept.com/staff/micah-lee/

Ge0rG: i complain all the time (in my head)

regarding http upload

steven, 1. Gajim doesnt blind trust, but every single user tells me i should implement it 2. you just exchange one verification for another, you dont want to verify the omemo fingerprint, and trust an pgp signature on it, but next you dont want to verify the pgp fingerprint, then you just trust some names on a list that maybe work in a newspaper

thats not how it works, if you want to be really secure, you have to put in the work

there is no magic solution how a computer can tell you that you can absolutly be sure that on the other end is Human X

at somepoint, someone has to check this in the real world

video

lovetox, I think steven mentioned that their company's employees verify their PGP fingerprints in real world

and then the next thing you have to realize is, that clients are not developed for 1% paranoid people

Wiktor, yeah so they know how this works, then they can do it with omemo fingerprints

all of your pgp signing theorys are way to complex to implement, its already hard to get omemo as is working in a usable way

yes, but for PGP once you sign a key the person can rotate subkeys freely and the trust is retained

with OMEMO there is no master key to hold device keys together

just clarifying what's the scope, I actually had an idea how to implement it outside clients using PGP but without modification from XMPP client developers using verified XMPP URIs (what basically is in the OMEMO QR code)

And? do you see anyone using pgp in xmpp?

Wiktor [21:16]: > with OMEMO there is no master key to hold device keys together And you have O(n*m) manual key management overhead

pgp has two components, identity verification and signing/encryption, pgp for xmpp as is today is used only for signing/encryption, not identity verification

Where n is your devices, and m the other users.

you already do M when you verify your users OMEMO keys?

the problem is you need to repeat it for every new device key

Thats the whole story of signal, no master key, its a feature that enables you easily add new devices

that is what makes it usable for the masses

now you want to "secure" that down to pgp levels

just use pgp

there is no way to use pgp identity verification in xmpp currently

pgp fingerprints are transferred in band in all pgp xeps I've seen

xmpp is just a transport protocol, everything pgp offers you can use

its like email in that sense, it transports the encrypted payload, you can verify around that with keyservers or whatever crazy construct you think up

verification of pgp keys can be done with QR codes like with OMEMO and with OpenKeychain, nothing uses that so bascially pgp in xmpp as it is now relies on server telling the fingerprints to clients, there is no paranoid mode like in OMEMO

but I think what steven proposed (as far as I understood) would be to use pgp keys that already have trust between them (bidirectional signing) to sign OMEMO device keys

and how do i get the public key to verify the sign?

dont tell me from a server :D

you get the fingerprint by scanning QR code, this is identical to OMEMO

see: https://github.com/open-keychain/open-keychain/wiki/QR-Codes

ok, so you dont want to scan the omemo qr code, because thats somehow to much work, thats why we sign the omemo key, then scan the pgp key that this was sign with

I don't want to scan omemo keys every time contact changes devices, pgp key is stable as it is the root of trust

to me this sounds like you just moved your problem and added complexity

and how does a user add a new device, where does he store his secret master pgp key?

you also don't really have to involve PGP to get the same thing right?

on the phone he just lost?

can't the device key you trust sign new device keys, and let you know about that?

this is just exactly what people do since 20 years with pgp

having a masterkey and singing sub keys

lovetox, usually PGP master keys are more protected than offline keys like OMEMO, e.g. my signing/encryption keys are on hardware tokens, master key is on an airgapped offline machine

lovetox, exactly

Wiktor, thats not usable for the masses

they dont store secret keys on hardware tokens

they get a new phone

log in, and want to chat

I meant something a little less strict, ie "trust any key I've trusted for x@x.com, and any new keys for x@x.com that one of my trusted keys have signed"

is verified omemo for masses? but it exists

thats what the signal protocol solved, thats why whatsapp is using this protocol for 1 billion people

so what you describe is not an issue with omemo, its a design decision to make it usable for the masses

if thats not secure enough just use pgp

and if the pgp UI in clients is not what you think it could be, work on that

instead of making omemo into something it was never designed to be

this is not an issue with "pgp UI" nor pgp as used for encryption, but if you say omemo should stay as close to signal as possible... okay

moparisthebest, yep, that sounds lightweight, there is an issue with revoking devices and tracking which device signed which one

uh, revoking is just "now my trusted key for x@x.com said not to trust this other key for x@x.com" ?

just have to be careful that the signed message going away alone doesn't revoke trust, since the server operator could pull that off

but it could also block the revoke message, I don't think there is anything you can do about that

it's at best a "my phone was stolen please don't encrypt messages to it anymore" switch

Yep, maybe the signatures and revocation can be embedded in XMPP QR codes as for OMEMO, that is transported out of band

Yes, stolen or unused anymore

yea that'd be pretty great

There is alternative to revocations - re-signing expiring signatures every N weeks or so

JWTs work like that... a little :)

then an evil server op can revoke keys though

trying to decide if that's a problem, I mean they can also just block messages

yeah

but putting these signatures in random messages would hide them :)