Re: [mdnsext] mDNSext features/requirements rollup

[RJ Atkinson <[e-mail]>]

On 15  Feb 2013, at 03:47 , David Farmer wrote:
> However, I will say that care needs to be taking to not proxy
> link-local (IPv4 or IPv6) addressing associated with services
> and only proxy GUA or ULA addressing for services off of a
> network segment with multicast or hybrid proxy of mDNS services.  

Off hand, I'm not sure exactly what set of rules/guidelines
ought to be promulgated, but I agree that this is a question
that this WG ought to examine -- and that some set of rules
or guidelines ought to be documented as part of any 
standards-track specifications.

> Furthermore, we need to define if a proxy should defend the
> name of a link-local only service on other segments to preserve
> naming coherency for any devices that may see the link-local
> only service and the potentially conflicting GUA addressed
> service of the same name on a different segment

I agree this deserves thought.  I don't have a firm view
myself just yet.

> Should an mDNS proxy also proxy the data path in the case of a
> link-local only addressed service as well?  

Also a good question that the WG ought to discuss and address.

> In the case of an mDNS proxy across a NAT Router then you
> would have to use the proxies outside address and NAT
> the Data path of the service, there are a whole bunch
> of corner cases here, BLECH!!

My intuition is that a NAT boundary (like an router interface
running E-BGP) is a natural indication of a natural 
administrative boundary.  

For a wide range of reasons, one normally ought not ("by default") 
proxy or forward mDNS packet across any natural administrative 
boundary.

Yours,

Ran