On Thu, Sep 03, 2020 at 06:13:41PM +0200, Florian Obser wrote:
> The point of this excercise is to work out if it's worthwhile to
> implement RFC 8806 "Running a Root Server Local to a Resolver" in
> unwind(8).
>
> We are trading latency for bandwidth. Lower latency is almost always a
> win for unwind(8) usecases. But the work if we fetch the root zone is
> not trivial either.
>
> The zone usually gets updated twice a day and is about 1.2MB. Due to the
> timing parameters unwind(8) would check every 30 minutes with a SOA
> query if a new zone is available:
> $ dig @k.root-servers.net +multiline +noall +answer . soa
> . 86400 IN SOA a.root-servers.net.
> nstld.verisign-grs.com. (
> 2020090300 ; serial
> 1800 ; refresh (30 minutes)
> 900 ; retry (15 minutes)
> 604800 ; expire (1 week)
> 86400 ; minimum (1 day)
> )
>
> The code complexity should be managable, all the bits and pieces are
> there in libunbound.
Curiously nobody reported unwind(8) doing tcp queries.
The histogram for udp looks like this:
0 50 3
50 100 3
100 150 1
150 200 1
200 250 1
300 350 1
700 750 1
1000 1050 1
1700 1750 1
1800 1850 1
2050 2100 1
5200 5250 1
9900 9950 1
35050 35100 1
--
I'm not entirely sure you are real.
