andir / local-nix-cache Goto Github PK

Site-local IPv6 multicast has a bunch of benefits compared to mdns. The most notably one is probably that it just works wherever IPv6 multicast routing works. This is true for most of the L2 segments on this planet. It might also work beyond a single L2 if the local network admin has enabled that. This gives the feature more reach than, say, simple IPv4 multicast in the local segment. This also means that we do not have to follow avahi anymore. We can produce richer messages and might not have to constantly search for peers. We can instead just listen for queries on our dedicated IPv6 address(es) and send out a single packet to query all neighbors.

Some time last year I've registered the following on the IPv6 multicast site-local registry:

FF0X:0:0:0:0:0:0:170
substitute X for any of the following:

         0  reserved
         1  Interface-Local scope
         2  Link-Local scope
         3  reserved
         4  Admin-Local scope
         5  Site-Local scope
         6  (unassigned)
         7  (unassigned)
         8  Organization-Local scope
         9  (unassigned)
         A  (unassigned)
         B  (unassigned)
         C  (unassigned)
         D  (unassigned)
         E  Global scope
         F  reserved

(from https://tools.ietf.org/html/rfc4291)

At first (and as default) we might just want to use FF05:0:0:0:0:0:0:170.

My initial idea is to just use DNS queries and have nodes respond to those (that are well formed) when they have a local (trustworthy) substitute for the query. Once the existance of such a path has been communicated we can fetch the narinfo & nar files via plain old unicast HTTP.