Policy-Discussion

[Philipp Gühring <pg AT futureware.at>]

Hi Peter,

> > How's tricks?  How's the IM world?  Did they ever get
> > their crypto sorted out?
>
> In the Jabber world we use TLS for channel encryption and have been
> making good progress with Duane on including the right ASN.1 bits (see
> Section 5.1.1 of RFC 3920) in domain certificates to use CAcert-issued
> certificates for secure server-to-server federation. Then we do SASL
> EXTERNAL for auth and Bob's your uncle.

I had some weird problems trying to get that running. Some servers still have 
serious troubles with TLS handling, and some clients don´t tell you, whether 
you are actually using TLS or not. (A friend of mine, who is a security 
specialist was a surprised, when I showed him, that his supposedly encrypted 
connection to his Jabber server was plaintext instead. He had activated "use 
SSL when available" in the configuration, but it didn´t worked, and he didn´t 
noticed.)  Things like that make you loose the trust in Jabber as a whole.

> Of course, if you don't trust the servers (and naturally you shouldn't)
> then we need end-to-end encryption, but we still haven't worked out a
> commonly-used system for that. :(  (Some clients do OpenPGP, none do the
> S/MIME stuff in RFC 3923, and the OTR-like protocol in JEP-0116 is still
> experimental.)

Then I suggest the following for you:

You should start a "Jabber compliant" campaign. Setting up a small 
test-center, and verifying the Jabber protocol compliant-ness and certain 
usability requirements, before a product can call itself "Jabber compliant". 
Define some minium "quality of service" that a product has to support, to be 
officially Jabber compliant.
Create a automated test-suite that helps you testing it.

If you don´t do it, the Jabber world will consist of interoperability issues, 
 
and the users won´t be happy about Jabber in the end, having too many issues 
with each different software package.

(Take a look at PKCS#11 for example, to see how bad non-quality controlled 
infrastructure can totally cannibalize the given standards)

Best regards,
Philipp Gühring