[dns-operations] how common is a 66-record answer section, i wonder?

[Paul Vixie]

> > i realize that .rhosts is dead, but there was a time when being able
> > to check gethostbyname(gethostbyaddr(getpeername(s))) == getpeername(s)
> 
> 	Which only required a single PTR record.  Adding extra PTR
> 	records usually meant adding extral lines to .rhosts as
> 	only the first PTR return was checked.

having personally fixed the rexec library to check multiple PTRs, i know.

> > was valuable, and being able to enumerate all of the owners of A RRs
> > that had the same rdata was therefore valuable.
> > 
> > it's still in common use for anti-spam MTA's.  postfix has an option
> > for "don't allow e-mail from hosts who don't have PTRs"
> 
> 	This doesn't require multiple PTR's.

it does if the host you're coming from has multiple names it can HELO from.


conceptually this is an expansion of:

	     EXAMPLE.VIX.COM.	  	  IN	A	128.45.1.1

	     1.1.45.128.IN-ADDR.ARPA.	  IN	PTR	EXAMPLE.VIX.COM.

to:

	     EXAMPLE1.VIX.COM.	  	  IN	A	128.45.1.1
	     EXAMPLE2.VIX.COM.	  	  IN	A	128.45.1.1
	     EXAMPLE3.VIX.COM.	  	  IN	A	128.45.1.1

	     1.1.45.128.IN-ADDR.ARPA.	  IN	PTR	EXAMPLE1.VIX.COM.
					  IN	PTR	EXAMPLE2.VIX.COM.
					  IN	PTR	EXAMPLE3.VIX.COM.

and everything that was designed to work with the former should dependably
work with the latter.  and a lot of small things, like me, now depend on it.

it isn't always possible, or desireable, to use a CNAME for EXAMPLE2 and
EXAMPLE3 as in the above example.  multiple PTR rdatas are how to maintain
A/PTR symmetry in such cases.