commit 791a1e246fac978b46503b7e757eb9eea4edf049
parent 624b583816a8915b8179365b9354caa19f275354
Author: Daniel Barlow <dan@telent.net>
Date: Mon, 29 Dec 2025 12:57:45 +0000
more mdns notes
Diffstat:
| M | notes.md | | | 35 | ++++++++++++++++++++++++++++++++++- |
1 file changed, 34 insertions(+), 1 deletion(-)
diff --git a/notes.md b/notes.md
@@ -186,8 +186,22 @@ status: done except that (c) probably the code is sucky;
** don't answer queries if the answers in the query already include what we'd send
+mdns is so fricking noisy already I don't know if this is worth the extra effort
+
** set the CACHE_FLUSH thing when needed
+should be set on unique records but not shared
+
+"all the records with that name, rrtype, and rrclass are conceptually
+under the control or ownership of a single responder"
+
+so this would be the Bike SRV and TXT and everything they reference
+but not the PTR records because other devices could also legitimately
+have those
+
+
+
+
** do we need to do NSEC?
** handle QU and legacy queries
@@ -208,8 +222,27 @@ status: done except that (c) probably the code is sucky;
// log::info!("IT ME");
}
+5.4 Questions Requesting Unicast Responses
+
+- top bit in the class field of a DNS question is the unicast-response bit
+- send repsonse unicast, unless the record has not been sent recently
+ ("within one quarter of its TTL")
+- also for queries received fom unicast
+ - but verify the source is local lan
+ - and the port is 5353 (otherwise see "legacy")
+
+
+6.7. Legacy Unicast Responses
+
+if source port != 5353
+
+- respond via unicast to source
+- query id and question
+- don't set cache-flush bit
+- ttl <= 10s
+
+
-mdns is so fricking noisy already I don't know if this is worth the extra effort
** compress responses
