On Sun, Mar 14, 2021 at 9:16 PM <fujiw...@jprs.co.jp> wrote: > Dear DNSOP participants, > > Thanks very much for good comments for > draft-ietf-dnsop-avoid-fragmentation. > > These are my proposal of Section 3.3. Default Maximum DNS/UDP payload > size. > > I'm not sure what to do with "MAY, "SHOULD", or "MUST", > so please give us your opinion. > > If it is acceptable, I will submit -05. > > --------------------------------------------------------------------------- > 3.3. Default Maximum DNS/UDP payload size > > Fujiwara-san, please consider the following addition at this place in the proposed text:
The default maximum DNS/UDP payload size MUST NOT exceed the calculated value based on the outgoing interface MTU. It MAY be configured lower than this value, based on knowledge of the operator's network vicinity and minimum link MTU of that vicinity, or operator preference using the table below. > Default maximum DNS/UDP payload size depends on the connectivity of > each node, it cannot be determined unconditinally. However, there > are good proposed values. > > Operators MAY select a good number from Table 1. > > +========================+==========+==========+ > | Source | IPv4 | IPv6 | > +========================+==========+==========+ > | minimal: RFC 4035 | 1220 | 1220 | > +------------------------+----------+----------+ > | Software developpers / | 1232 | 1232 | > | DNSFlagDay2020 propose | | | > +------------------------+----------+----------+ > | This document proposes | 1400 | 1400 | > +------------------------+----------+----------+ > | maximum: ethernet MTU | 1472 | 1452 | > | 1500 | | | > +------------------------+----------+----------+ > | calculate | MTU-20-8 | MTU-40-8 | > +------------------------+----------+----------+ > > Table 1: Default maximum DNS/UDP payload size > > Similarly, I have a suggestion for replacement text for the next sentence: However, operators of DNS servers SHOULD adjust the maximum DNS/UDP payload size at installation/configuration time, or at start-up time, accordingly. Clients SHOULD perform MTU validation* (rather than discovery per se) towards configured resolvers at start-up. Authority servers' operators SHOULD monitor for PTB messages, and passively collect TCP MTU values, and adjust accordingly. (* I have a separate suggestion for a mechanism for validating MTU, which I will send to the list separate from this.) > However, operators of DNS servers SHOULD measure their path MTU to > well-known locations on the Internet, such as [a-m].root-servers.net > or [a-m].gtld-servers.net at setting up the servers. The smallest value of path MTU is the server's path MTU to the Internet. Then, in this section, substitute "configured or adjusted" for "reported". The > server's maximum DNS/UDP payload size SHOULD be smaller than or equal > to the reported path MTU minus IPv4/IPv6 header size (20/40) minus > UDP header size (8). > > Sincerely, Brian Dickson
_______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
