XSF Discussion - 2014-03-06

"There is a lesson here. Standards are there to make your life easier." -- http://www.chriskranky.com/amazon-mayday-maybe-using-webrtc-cares/

I thought standards are there to prove I'm right, for all values of "right"

they're terribly useful for that, too

https://twitter.com/MaciejMusialik/statuses/441515074063466496

sadly too true

but fippo is fixing that, right fippo, right????

later today (-:

after talky finally has its turn servers

awesome

btw: there is a TRAM working group that is improving TURN

if you're not already paying attention there, I think we should

(TRAM working group at the IETF)

i'm paying attention

but my need for turn is mostly satisfied by draft-uberti-behave-turn-rest ;-)

And there's a TRAM session later today

Oh, I forgot about DANE.

And that's now :)

is TRAM today?

https://twitter.com/danyork/status/441503787560493056 <-- there was no space left

oi

too many conflicts

Zash, Indeed. Is it interesting?

I think, looking at the agenda, I've missed everything I meant to listen in on anyway.

m&m: or .. I might have been looking at yesterday

so nm

dwd: Interesting. But now it's semantics ^^

dwd: way to go

In my defence, I've been quite preoccupied recently. :-)

I'm vaguely regretting only turning up for Tuesday.

If timing had been different, I'd have thoroughly enjoyed turning up for the week.

This week wasn't convenient for me, I had to be home yesterday.

I'll just have to go to Hawai'i instead.

Haha

Zash, Someone's got to do the hard jobs, you know.

you clearly weren't at the admin plenary; Hawai'i is a terrible burden to go to

but all the important decisions will be made there

fippo, Put your expense claim in early.

so like it or not, you have to go

Right. Certificates for XMPP servers. Do folks still use startcom?

Folks do, yes.

I went for a Comodo cheap-but-not-free cert.

there's quite a few

How cheap is cheap, and why is it better than free? :)

I think my two certs were somewhere around £15.

StartCom are free, but revokation costs about £50, whereas it's free with me.

that is a good point

you want revocation to be cheap, in case you get compromised

So to some extent, I'm paying for a brand I trust to know what they're doing, and paying an insurance premium.

m&m, Right, I don't want to get compromised and *then* stung for a hefty fee.

"Starting from £41.95 per year"

This sound like significantly more than £15.

I think Startcom is free if you paid for a premium cert

https://www.namecheap.com/security/ssl-certificates/domain-validation.aspx

dwd: The text on https://www.namecheap.com/security/ssl-certificates/comodo/essentialssl.aspx, which is the £15ish one, seems to suggest it's single-domain-only (i.e. no MUC child). Sound about right?

Ah, yes. My two are single domain. But dave.cridland.net was covered, unlike StartCom.

FWIW, I think mine are the top one on that page.

I have one for cridland.im and one for dave.cridland.net

Ah, the £6/year ones?

Right. I think. :-) It was a while ago.

It's entirely unclear to me what the difference between PositiveSSL and EssentialSSL is :)

Ah, that one has unlimited reissues. Let me dig through and see what I have.

Murky.

Ah, so I have EssentialSSL certificates. I have a feeling they were on offer when I bought them.

So these have a "site seal", which I don't bother with, "mobile browser support", which I don't think means much, and unlimited reissues (ie, for compromise or whatever).

Getting two or three EssentialSSL certs seems a tad expensive.

£43/year or whatever. Almost worth going with a filthy wildcard at that price.

startcoms wildcards are $60 for two years it seems ✏

Hmmm. We could always see if we could persuade a CA or two to give XMPP folk a discount because we're so lovely.

Temptation to just get a couple of these £5/year certs is fairly strong. Although I don't see anything about the reissues on the pages.

Right, I think on compromsie you pay again.

But they have a "live chat" thing which has people who're knowledgeable about these things.

By which I mean they'll be able to tell you about revokation etc, not that they can tell you much of note about odd X.509 features.

Kev, Benefits of working for Isode - discover you now know more about X.509 than most CA employees purely by osmosis.

Hahaha

you can get free wildcard certs for opensouce projects from Globalsign.

and startcom charges for all revocations, except for their EV certs http://www.startssl.com/?app=25#72

dwd: you almost say that like it's a good thing

https://www.globalsign.com/ssl/ssl-open-source/

m&m, If only it was.

Tobias: right. "premium" (-:

simon: Ta. This is for my own server, rather than an OSS project.

Kev: with a free cert from startcom you would be able to cover muc.<domain> and <domain> since they always fill out the altname too

Right. That was how this conversation started :)

Although if I wanted to bring channels.doomsong back to life, I'd need a third domain :)

or an additional cert just for that

Tobais: I don't think that would work - at least not from startcom - they notice that you are trying to get another free one for the same domain.

simon, i didn't mean from startcom...right..they'd probably notice :)

You can certainly request foo.domain.tld and bar.domain.tld at StartCom.

simon, What's actually in a StartCom cert these days?

no matter what you put in your generated cert, they remove it all and put <domainname> and a hostname portion that you can select.

Right, I remember that, but what's in the Subject, and what SANs are in it?

for one of my keys, I have roughly this:

Subject: description=7u4x3xy29u755HYu, C=NL, CN=owncloud.ik.nu/emailAddress=hostmaster@ik.nu

X509v3 extensions: X509v3 Basic Constraints: CA:FALSE X509v3 Key Usage: Digital Signature, Key Encipherment, Key Agreement X509v3 Extended Key Usage: TLS Web Server Authentication X509v3 Subject Key Identifier: 72:CE:E6:0C:5F:D5:EA:54:BB:F9:A8:42:28:AF:F9:DE:60:DA:9F:F5 X509v3 Authority Key Identifier: keyid:EB:42:34:D0:98:B0:AB:9F:F4:1B:6B:08:F7:CC:64:2E:EF:0E:2C:45 X509v3 Subject Alternative Name: DNS:owncloud.ik.nu, DNS:ik.nu X509v3 Certificate Policies: Policy: 2.23.140.1.2.1 Policy: 1.3.6.1.4.1.23223.1.2.3

That's a regular one, not one I requested for XMPP.

Those have othername fields in the SAN

intosi, Right, Sodium will tell you what those are, mind. There's a tool I miss having around.

dwd: I think if you'd seen the 16.3 MLC (which doesn't exist yet, but we know what's coming), you'd miss having that around as well :)

16.2 MLC is really rather good, mind. What was the last version you saw?

Kev, dwd: :D

Kev, A R15.X, which was certainly getting there.

Ah. Worlds apart :)

What was the deal with special XMPP certificates a few years ago? What was different about them?

simon: They had the right SANs in.

ah

simon, As I recall, they listed SANs, but had some funnies around the sRVName SANs they used.

dwd: I think the XSF certs were correct, IIRC.

Back when we had an ICA.

Why would a normal alt-name not work?

simon: It's not 'not work', it's just that certs should be specialised for the service they're protecting.

simon, It would, but older servers were fairly restrictive in what SANs they used.

simon, ALso, there's no such thing as a "normal alt-name". :-P

simon, The Subject is a DN, originally meant to be your entry in the global X.500 directory. The Subject could have alternative names (added in v3), which are all typed. dNSName is the hostname type, and otherName is an extendable type where both sRVName and xmppName live.

Which openssl conveniently refuses to display ;)

dwd: you really need to start a CA

simon: we somewhat jokingly discussed this a feww years ago, but I think the rough consensus was that running it would be entering a world of pain.

In practical terms, yes. Although in technical terms, Sodium CA makes all this rather easy.

simon, I've also contemplated a CA based around leap-of-faith verification before.

I do have my own CA I use 'internally'.

Kev, Technically, yes; I think the objections were more political ones.

Right.

Running your own CA isn't the world-of-pain part. Getting your CA accepted as a trust anchor in major browsers is.

the pain of getting it in the major OSes and browsers

If it was a CA for XMPP, you don't need to do that.

you'd need to provide guides for all XMPP servers how to add your CA to the trusted ones

Kev, That's true, in some respects. Though you need to get it in all the XMPP implementations.

https://bugzilla.mozilla.org/show_bug.cgi?id=647959

I got an Ubuntu update pushed today that removes CACert as a CA

saywat

ubuntu only or did debian get rid of it too?

dunno

ralphm: What Ubuntu version?

13.10

https://bugs.launchpad.net/ubuntu/+source/ca-certificates/+bug/1258286

http://changelogs.ubuntu.com/changelogs/pool/main/c/ca-certificates/ca-certificates_20130906ubuntu2/changelog

looking at the full change list of that ticket reveals it was backported to lucid, precise, quantal, raring

I don't see that update in precise

-proposed

?

Aha

Are they actually really removing it completely?

As opposed to not having it enabled by default.

"No longer ship" seems to suggest they have removed it completely.

ralphm should be able to confirm.

ralphm@waar:/etc/ssl/certs$ ls | grep -i cacert spi-cacert-2008.pem

http://packages.ubuntu.com/trusty/all/ca-certificates/filelist

It doesn't list /usr/share/ca-certificates/cacert.org anymore.

right

Fun

organizationName: Software in the Public Interest

:(

where?

Where what?

> 13:57:29 Zash> organizationName: Software in the Public Interest

Sooooo much lag on the IETF wifi

That was re: ralphm> spi-cacert-2008.pem

Seems like Fedora, Redhat and Suse are also not too keen on CACert inclusion

It's likely that it's only in Ubuntu because it's in Debian

Hardly anybody was keen on that, mostly because it didn't pass the audit. Of course, recently the found vulnerability and subsequent lack of revocation of the ca key did not improve that. http://www.reddit.com/r/technology/comments/1qj1tz/http_20_to_be_https_only/cddfmz0?context=1 (fourth para)

I think CAcert.org themselves aborted auditing while waiting for some changes to be made.

intosi: great paragraph / nice background.

scribing to http://etherpad.tools.ietf.org:9000/p/notes-ietf-89-kitten?useMonospaceFont=true

BTW thanks for everyone who came to XMPPUK on tuesday. Hope everyone had a good time/got something out of the evening.

Lloyd: even though I wasn't there, thanks for doing that!

yes, thank you!

Indeed. It was a very good meetup, thanks!

Lloyd: Yes, thanks. And plenty of pizza :D

Well, golly. They're asking for a CSR. I guess I should work out what to put in it.

dwd: Any idea if they pay any attention to what you put in it, or if they're just going to trample over and I don't need to bother?

Most CAs will replace it with the information they have on record anyway, because that's the only info they verified.

It's either that, or requiring you to send proof of identity with each CSR.

https://www.dropbox.com/s/et86sczq4h76r4u/Screenshot%202014-03-06%2015.45.44.png whaaaaaaaaat?

They want to hold your hand installing the signed cert?

I assume it's to deliver it in an appropriate format, but I still found it slightly surprising.

Oh, or that possibly, yes.

Kev - which register is that and which register do I need to avoid?

Some of those products are ancient!

Comodo, via Namecheap.

But it seems to be to guide you to installation instructions, so it's fine.

Kev, I don't think they used anything but the public key.

And there, prototype s2sin DANE.

Zash: excellent. Looking forward to a new weekend project.

But I'm back to it being a race condition :|

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

zash: the "no port" problem sounds familiar....

Right, SRV-like is how (IIRC) dane-esmtp works, isn't it?

-draft dane-smtp

Zash: "SMTP security via opportunistic DANE TLS", Viktor Dukhovni, Wesley Hardaker, 2014-02-14, http://tools.ietf.org/html/draft-ietf-dane-smtp-with-dane-07.txt

That one?

dwd: I spent yesterday searching for anything existing on s2s client auth, found only this thread: http://www.ietf.org/mail-archive/web/dane/current/msg05110.html

I really disagree with "the stream is not an XML document" viewpoint

It's not, as a whole.

Yes, it is

It has an opening tag and an ending tag

It has multiple opening tags, and one ending tag.

No, it doesn't

One opening, one closing

You are confusing it with the other unfinished streams that went before

Throwing away the state each time you restart is not elegant from the XML PoV.

Except that you restart writing out the document without closing it.

dwd: But dane-smtp and dane-srv are meant to be in harmony.

MattJ: No, you're confusing my use of 'stream'.

So what? Any XML parsing lib lets you throw the parser away and start another

Kev, then say "connection"

That's...actually very not true :)

Kev, e.g. ?

The number of libraries I had to go through in Java before I found one that let me work on an incomplete stream without waiting for the end was depressing.

But this is orthogonal to the stream restart stuff.

That's not quite the same thing

Such libs are clearly not applicable to XMPP :)

So you need a SAX parser, we knew that already.

That alone is not enough :)

Although this is more a comment on the sad state of Java XML parsers.

And doomsong.co.uk finally has an A in the observatory. How nice.

(Just so long as no-one looks at the subdomains)

Let's talk about export ciphers

MattJ: They're disabled.

Welcome to the club Kev.

Kev, on jabber.org?

Oh, no.

Not there :)

Right

To continue Tuesday evening's discussion...

Is the suggestion that jabber.org would be breaking the law to disable them? (seems ridiculous to me)

or is the argument that people might be using software that only supports them, and we must allow that?

I heard Intosi claiming the latter.

From what Kev has said in the past I assume the latter is the case, so I don't know how legality came up in conversation

Well, I suppose xnyhps making a US-centric statement :)

I can imagine that those old clients using old ciphers are probably unused / installed at one point and sitting in a windows95 taskbar sucking the odd cpu cycle.

kill the zombies.

Indeed, I honestly think that providing people using such software with insecure service is doing nobody any good

+1

MattJ: The claim was made that the old export cypher laws were no longer relevant. This isn't true.

Agreed

This isn't related to j.org's choice of suites.

Ok, fine

(and good)

Did Jabber.org take part in the last test-day?

Yes.

How did it work out?

Number of S2S dropped a lot, I think, but I didn't check.

The main complaints were Google-hosted domains.

I'm quite encouraged - we started with 2% forced encryption on s2s traffic - that's almost up to 15% now.

?

https://xmpp.net/reports.php#starttls

sorry - been a long day. Tried to kill -9 <file> a few moments ago.

MattJ: Now I already made it obvious I don't know much about these laws, but don't they cover only exporting software *itself*?

So not offering a service?

xnyhps, correct

Also not applicable to open-source software (i.e. OpenSSL, GnuTLS)

I would presume bundling such software with commercial software may be problematic though

(I did try to read the Wassenaar agreements a week or two ago, but couldn't get further than a couple of lines with all the legalese.)

Presumably any site offering an HTTPS connection with strong ciphers would be in breech of whatever agreement.

Well they are all outdated, and don't really match up with the way software, services and the internet work nowadays

Kev: well done.

The cipher export laws in the UK only affect stuff for which the source code is not available (ie, non-open-source), and they're an implementation of EU directives.

The problem isn't so much the EU directive, but the fact they're enforced by a bunch of civil servants who're out of their depth.

Well, if the directive still technically forbids the export of >56 bit symmetric and >512 bit asymmetric encryption, then I'd say it is a problem.

Time to lobby our MPs? :)

But jabber.org has ciphers enabled that are even weaker than this laws would allow.

*this law

xnyhps, The way it's implemented in the UK is that exported software must have the means to disable "non-export" ciphers, and that those ciphers are off by default. Basically, Isode's licensing keys are largely about implementing this requirement.

dwd, meaning it's fine to ship the software with a way to turn strong ciphers on?

(YANAL, I know :) )

The interesting grey area is that it's the use of encryption, and not the implementation thereof, so even using platform crypto is possibly problematic.

MattJ, Right, that's what Isode do.

If anyone cares about this stuff, https://www.gov.uk/export-of-cryptographic-items

Now I'm curious which of the conditions in the Cryptographic Note Isode's stuff doesn't satisfy.

Which ones do you think it does?

Sold without restrictions sounds likely. Not easy to change is somewhat inherent to it being crypto. Designed to be installed by the user without support, I don't know. Providing details on request, why not?

#1 is arguable, #2... do config files count? I'd say so. #3... my guess is that Isode selling software without support is unlikely and #4 can be complied with by anyone

MattJ: You can't replace the Isode crypto by changing config files.

MattJ: ‘The cryptographic functionality cannot easily be changed by the user’ means that the manufacturer has taken reasonable steps to ensure that the cryptographic functionality in the product can only be used according to their specification. That suggests that if they define the config files as the specification, they're fine.

Fine

xnyhps, "cannot easily be changed by the user" implies config files are not fine.

<!--This is an automatically generated configuration file and must not be manually edited.-->

(From an M-Link config file)

simon: Does that count as not letting users edit it? :)

:P

In any case, it looks like I wasn't right; open source would be fine, Isode's stuff would need a license, but you could probably manage to ship a simple closed-source XMPP server under the rules too.

Well, it probably won't fly to just specify "users can do EVERYTHING with this", but it suggests that you can.