XSF Discussion - 2014-03-17

dwd: ping. are you still overloaded with work?

Zash has DANE working for Prosody now. It looks like Tigase has something on the horizon: https://projects.tigase.org/issues/1626. Anyone know about Ejabberd or MongooseIM or Openfire?

Need to write a proper DANE+XMPP draft

Zash: Do you have some setup instructions lurking anywhere?

simon: For the Prosody DANE plugin, DNS setup or bot?

both*

Zash: would be acceptable to put the XMPP-specific parts of DANE into draft-ietf-xmpp-dna, or does it really require a standalone draft?

The plugin looks pretty straight forward. But for the DNS setup. All the guides I've read are quite complex and assume a massive rollout... Well it would be nice to have a simple setup guide. Does such a thing exist?

m&m: Would probably be fine to have it in DNA-DANE

Zash: patches welcome (-:

Maybe, once I recover from the meeting I had this entire weekend.

ouch

working on that is third on my TODO

DNA-DANE-POSH maybe

we still need to define what DNA is actually ;-)

http://dictionary.reference.com/browse/DNA

My code is currently diverging from the DANE-SRV stuff

Zash: how does it diverge?

It looks up _xmpp-server.example.com IN TLSA instead of _$port._tcp.your-srv-target.example.com

I doubt that TLSA for _xmpp-server.example.com would get published

Hm?

what's your thinking on using xmpp-server instead of _$port._tcp.your-srv-target.example.com?

simon: https://www.zash.se/dane-s2s-client.html

I love it when someone can point to a webpage to answer questions!

the problem here is that it assumes a level of micro-coordination between the owning domain and the hosting provider that is only really possible for stand-alone servers

Basically, it's easier, especially for validating incoming connections.

m&m: CNAME :P

/-:

\-:

requiring CNAME to deploy seems like a Really Bad Idea™.

and again requires the owning domain and hosting provider to be in sync on DNS updates

It's bad enough getting people setting up SRV records... (SRV what?)

also, is this format for TLSA even valid?

ralphm: it's not prohibited, but it's not documented anywhere

simon: strongly disagree. I wish SRV was used more. Like in HTTP.

RFC 6698 documents using _$port._tcp.hosting.example.net

ralphm: The qname has nothing to do with the TLSA format, that's just an example convention.

ralphm: wishing something were so and reality are often in conflict (-:

Ralphm: I agree- wish they were used more too. But I have to explain why they are awesome every single bloody time.

and how 5222 and 5269 are not "proprietary ports"

simon: write a web page to point to.

m&m: Incoming s2s connection, I wanna validate the client cert without doing over 9000 lookups.

ralphm: at that point, I might as well use .well-known and host metadata d-:

And SRV records might not even point to servers doing outgoing connections.

m&m: I mean a web page explaining SRV

Zash: I understand your desire, but I think you're going to have to suck it up and deal with the extra lookups

You're going to do a million lookups anyway, because of DNSSEC.

+1 sucking it up

m&m: +1

intosi: The dns library handles that for me.

Zash: it's not a ton of lookups, but n*2 (n == count of SRV RRs)

well, 3*n if you do both A and AAAA and TLSA

One million lookups http://img2.wikia.nocookie.net/__cb20090506185112/uncyclopedia/images/f/fa/Dr_evil_one_million_dollars.jpg

And you probably will want to look up / have looked up the SRV records for the domain in question.

but really, anything that relies on the source domain holding information on the keying material severely constrains hosting deployments

intosi: Not for incoming connections with bidi

I don't want to promote any non-standard use of existing DNS RRs. There are a few to choose from for pointing to things, all with their own uses. TXT, CNAME, SRV, TLSA, PTR. Please, just use them appropriately.

I remember a recent discussion with simon on this for BC.

Go PTR!

I remember a recent discussion with DANE people at IETF that this made more sense than the _port way.

if it wasn't Viktor Dukonvni or Wes Hardraker, then it probably wasn't with people actually deploying things (-:

where's that email

http://www.ietf.org/mail-archive/web/xmpp/current/msg03231.html

m&m: that's one of the most pleasant to read RFCs. Nice job btw.

Zash: /sigh … this is after spending about 2 minutes trying explain XMPP S2S to someone unfamiliar with XMPP

woooosh

for actual XMPP clients, this might be ok, although I still doubt the actual deployability of true client certs

for servers, I think you'll have to do the SRV and TLSA lookups, or you could be getting false negatives

particularly for hosting providers that have separate certs for each individual end-point

simon: we try (-:

false negatives?

for instance, let's say hobbiton.example is hosted at a large provider with 6 s2s end-points(im1.middle-earth.example, im2.middle-earth.example, im3.middle-earth.example, im4.middle-earth.example, im5.middle-earth.example, im6.middle-earth.example)

you get an incoming connection from one of these, and you do a IN TLSA _xmpp-server.hobbiton.example

I get 6 records back, find at least one matching and then I'm happy.

if that resulted in 6 different TLSA RRs, then you'll probably be fine unless im3.middle-eath.example had to rotate keys

but if you relied on your CNAME, well things get a lot messier

Because?

if you did the SRV then TLSA lookups, you'd get a whole lot closer to reality

CNAME to multiple records are messy

Is it?

that's what I've been told at least (-:

plus, you're requiring the owning domain and possibly hosting provider to make sure extra records are in sync, which increases the likelihood of failure

doing the same SRV + TLSA dance for incoming connections as outgoing minimizes the number of records operators have to publish (reducing the bugs)

also, you can't point to CNAMEs with SRV

ralphm: you can have CNAME _xmpp-server._tcp.example.com

I'm sure that goes for other record types, too

ralphm: Not relevant.

ralphm: it also turns out the vast majority of resolvers don't really care where the CNAME/DNAME records are in the chain )-:

CNAMEs only exist because there wasn't anything better

the other thing with doing the SRV + TLSA dance on incoming is that you'll almost always have cached records for the outgoing (or almost always have cached on your outgoing before you got the incoming)

Ralphm: kinda like MX too?

simon: well, yeah, MX is just e-mail specific SRV

ralphm: kind of (-: It turns out to be more complicated than that, mixing MSA (Mail Submission Agents) and MTA (Mail Transfer Agents)

I'm still watching this bug: https://bugzilla.mozilla.org/show_bug.cgi?id=14328

one can have hope...

m&m: arguably, e-mail now has different services for submission and transfer

the fact that people still submit on 25 is, well, backwards

(so _smtp vs. _submission)

m&m: almost always, unless you're the receiving server. but then you need to outgoing dance soon anyway ;-)

fippo: that's my point (-:

I thougt people agreed with this already..

Quick question before send out an email to the list about the <potential> DNSSEC grant: Who is it from?

um?

the Internet Society are the folks we'd ask to fund this

Thanks. That's what I needed.

Doing my board task from last week - email about how we get most bang-for-buck from it.

Simon: thanks!

Have you planned an article for the next Security Test Day? "March 22, 2014 - third test day"

Neustradamus: I'm working on that.

I'm going to write an article for Russian speaking audience. You may know, we have a pretty large community, including ejabberd developers, but still no one takes part in the test days. I'll try to fix this. Just FYI:)

Hi Santiago26. Which are the big XMPP servers in Russia?

Simon: :)

Hi. Ya.ru (by yandex.ru search engine), qip.ru (QIP IM) — millions of users, hundreds of thousands online daily …and jabber.ru is the godfather, but it has not so many users

Santiago26. Thanks. Do you have any way to track down the admins for those servers? It would be great to shoot them a private message from either you or myself with a quick heads-up.

also wondering why qip.ru runs a quite old version of ejabberd

I think that yandex.ru admins are not accessible for me (us), qip.ru have some problems with their bosses, but we are discussing this idea, and i've tried to discuss Manifesto with zinid and ermine from jabber.ru and have no answer.

Tobias: It is custom version of ejabberd, forked at 2.0.3, I think.

Santiago26, ahh..i hope it's well maintained :)

http://blogs.adobe.com/standards/2014/03/18/the-business-of-standards-part-1/ -- true, true

fippo, nice read

Tobias, fippo, yes, but I can't see it separate from the context this is posted on

Adobe isn't necessarily on the forefront of standards development.

ralphm, wasn't it behind standardizing PDF at ISO?

but yeah..that's an old one

ralphm, what context are you referring to exactly?

Tobias: well, mostly that it is posted by Adobe

ah..ok

and also that the context is a bit missing, I suppose

why it he talking about standards, really?

right...that i wondered about too..i first thought i missed some WebRTC news or so :)

i mean while HTML5 and WebRTC are taking the biggest chunk out of flash over time, are they active in the WebRTC development or mostly just HTML5?

About that, couldn't do our weekly hangout because Hangouts is having issues. Used talky this time. Nice.

I haven't conciously noticed Adobe being active in WebRTC, but I am not following in minute detail.