Permalink
Browse files

mdnsd should attach EDNS0 on its own, and should not forward it -

should decide if it should attack EDNS0 or not, by itself.
  • Loading branch information...
1 parent 52668cc commit 17715932ec8f7f5b6b6efe34f715b8037d4002bd itojun committed Jun 30, 2001
Showing with 13 additions and 3 deletions.
  1. +4 −2 kame/kame/mdnsd/mainloop.c
  2. +9 −1 kame/kame/mdnsd/mdnsd.8
@@ -1,4 +1,4 @@
-/* $KAME: mainloop.c,v 1.67 2001/06/29 19:26:41 itojun Exp $ */
+/* $KAME: mainloop.c,v 1.68 2001/06/30 05:57:33 itojun Exp $ */
/*
* Copyright (C) 2000 WIDE Project.
@@ -39,7 +39,9 @@
* - negative cache on NXRRSET reply on query timeout
* - attach additional section on reply
* - random delay before reply
- * - as querier, retry by TCP/EDNS0 on truncated response
+ * - don't just relay EDNS0, decide EDNS0 independently from the querier
+ * - as querier, retry without EDNS0 on FormError
+ * - as querier, retry by TCP on truncated response (EDNS0??)
* - multiple replies
* - how long should we wait for subsequent replies?
* - conflict resolution
View
@@ -1,4 +1,4 @@
-.\" $KAME: mdnsd.8,v 1.39 2001/06/29 19:39:53 itojun Exp $
+.\" $KAME: mdnsd.8,v 1.40 2001/06/30 05:57:33 itojun Exp $
.\"
.\" Copyright (C) 2000 WIDE Project.
.\" All rights reserved.
@@ -303,3 +303,11 @@ sends ICMPv6 node information queries to link-local all node multicast address
We should be using NI group address
.Pq Li ff02::2:xxxx:xxxx
instead, however, it is not useful right now due to deployment issues.
+.Pp
+.Nm
+relays EDNS0 pseudo RR from a original querier,
+while it should decide EDNS0 independently from the querier.
+Imagine cases where the original querier supports EDNS0,
+while multicast/unicast nameservers
+.Pq which are the destination of the relayed packet
+do not.

0 comments on commit 1771593

Please sign in to comment.