Re: [homenet] Understanding DNS-SD hybrid proxying [was: Firewall hole punching]

[Markus Stenberg <[e-mail]>]

On 23 Nov 2016, at 21.45, Juliusz Chroboczek <[e-mail]> wrote:
>>>> - ohybridproxy (only really scalable and sensible IPv6 rdns source that
>>>> I am aware of, given nodes talk mdns)
> 
>>> Noted, thanks for the opinion.  I still don't understand how it works (who
>>> gets port 53?  how are data from multiple links merged?), but I intend to
>>> do my homework.
> 
>> I give dnsmasq port 53, and then have it forward queries for .home
>> (chuckle) and my IPv4/IPv6 reverses in .arpa-land to 127.0.0.1:54 where
>> ohp listens on my routers.
> 
> Ok, makes sense (except for the choice of 54).  Two more questions:
> 
>  - who merges data from multiple links?  (I'd wish that the hybrid
>    proxies compute a minimal spanning tree and perform peer-to-peer
>    magic, but I suspect you're generating a config file dynamically
>    and restarting dnsmasq whenever the set of hybrid proxies changes.)

There is no need for merging, there are only few zones. They are all in DNS-SD browse/legacy browse path, and also in DNS search path. The configuration is actually static in my case. The benefit of merging is limited as there are only few subnets.

>  - who speaks mDNS?  The Hybrid proxies?  Or do they communicate with
>    a dedicated mDNS speaker?

ohybridproxy is not a mDNS implementation. I did one once ( https://github.com/fingon/hnet-core ) and later on decided it was a bad idea. ohp uses patched version of Apple’s mdnsd  ( https://github.com/fingon/mDNSResponder ) for heavy lifting.
I haven’t recently checked, but at least at the time Apple’s UNIX version of mDNSResponder was essentially broken but I fixed the few bugs that bothered me when developing ohp.

Cheers,

-Markus