Re: [BEHAVE] draft-xu-behave-nat-state-sync-00

[Simon Perreault <[e-mail]>]

Xu Xiaohu wrote, on 2009-11-22 22:51:
> Could you or anybody else provide me some protocol specification for the pfsync?

Here's what Ryan McBride had to say about this:

> There is no formal protocol description.  It's been asked for before,
> but I've never been really interested in producing such a document.
> Several people have offered to write it up as a draft, and I'm willing
> to help review and edit a draft that someone else puts together, but
> somehow people never manage to produce anything for me to look at :-)
> 
> This links below are probably the best description of the protocol
> as it currently exists:
> 
> http://undeadly.org/cgi?action=article&sid=20090220014805
> http://undeadly.org/cgi?action=article&sid=20090222155558
> http://undeadly.org/cgi?action=article&sid=20090226064146
> http://undeadly.org/cgi?action=article&sid=20090301211402
> 
> Anyone designing a sync protocol should look at the discussion in these
> pages, because we've done a lot of learning with pfsync over
> the years, and what dlg@ has done recently is fantastic.
> 
> However, while the PFSYNC protocol itself may be a canditate for
> standardization, the message payload layout is most certainly not.
> They're tightly tied to the implementation of PF, so unless you simply
> define the messages to be sent and left the content up to the
> implementation, it's going to be an ill fit for other code. The protocol
> is pretty flexible about message types though, so it would be possible
> to specify a bare-minimum "required" message, and allow space for
> vendor-specific messages.

Simon
-- 
DNS64 open-source   --> http://ecdysis.viagenie.ca
STUN/TURN server    --> http://numb.viagenie.ca
vCard 4.0           --> http://www.vcarddav.org