[Council] Message threads

[Chris Mullins]

We use GUID's (off the top of my head) in:

1 - Message Threads
2 - ID attributes for IQ packets. We may eliminate this soon, as it sure
makes trace look ugly.
3 - Stream ID's during SASL (s2s)
4 - Dialback open stream responses 
5 - UserId's for anonymous users
6 - Resource ID's for unrequested resources during resource binding
7 - PubSub Node ID's
8 - PubSub Subscription ID's
9 - Unspecified Var names in DataForms
10 - the SASL CNonce value in Digest MD5 auth 
11 - the SASL Nonce value in Digest MD5 Auth

There's probably a bunch more. 

The original GUID spec (years and years ago) used a MAC Address in the
algorithm. In this more-secure-world that algorithm has been discarded
in favor of one that doesn't have identifying information in it. 

--
Chris Mullins


-----Original Message-----
From: council-bounces at jabber.org [mailto:council-bounces at jabber.org] On
Behalf Of Ian Paterson
Sent: Thursday, January 11, 2007 6:00 AM
To: XMPP Council discussion list
Subject: Re: [Council] Message threads

Chris Mullins wrote:
> We use big-ugly-GUID's all over the place. These are 
> (in essence) big ugly random numbers. 
>
> Lots more info at:
> http://en.wikipedia.org/wiki/Globally_Unique_Identifier
>   

Thanks. From reading that, it seems experience has shown we have to be 
very careful about the code we use to generate GUIDs. Without analysing 
the code (or using widely-trusted open-source code) we can never be sure

that no (personally identifiable) information is encoded in them.

In fact IMO it's worth adding a Security Consideration about that to 
XEP-0201 (Threads).

Chris, you say you're using them "all over the place". Are we using 
GUIDs or random numbers in any more XEPs (other than the e2e XEPs)?

Perhaps we should write a more general "Security best practices" XEP 
(after we've got the e2e XEPs to Draft)?

- Ian