[OTR-dev] Queuing of messages before the connection is established

[Paul Wouters]

On Tue, 8 Feb 2005, Evan Schoenberg wrote:

> OPPORTUNISTIC policy, and have already exchanged fingerprints.  Isn't 
> the message queuing supposed to handle the case where:
> - we are  having a secure chat
> - he quits and relaunches his client.  I'm still in a secure chat, with 
> him, as far as my client is concerned.
> - he sends me a message in plaintext
> 
> Shouldn't it re-negotiate the secure connected before he sends, if 
> possible, and then send his message, rather than sending his message, 
> telling me the message was not encrypted, and then handshaking?

Opportunistic is not an advertised property (yet?), so when one end restarts,
it can only either 1) try OTR blindly, or 2) send first msg with the whitespace
probe.

However, you are right that a changed behaviour would be nice.  One could argue
that we would like to try the reverse:

If we have talked OTR to someone before, and we just come up and do not know
the OTR status of that person, try OTR. Or make it more general, and say try
using the last state used. Though I'd prefer it to try OTR if it knows the
other end has spoken it before. I'd rather err on the presumed security, then
send too much unencrypted messages.

Paul