nanog mailing list archives

RE: lame delegations


From: woods () weird com (Greg A. Woods)
Date: Mon, 21 Aug 2000 13:31:55 -0400 (EDT)


[ On Monday, August 21, 2000 at 09:16:53 (-0700), Karyn Ulriksen wrote: ]
Subject: RE: lame delegations

Unless I misunderstand what you mean, my version of BIND (8.2.2p3)
doesn't do that.

    $ host -a 2.254.92.204.in-addr.arpa 
    2.254.92.204.in-addr.arpa       PTR     most.weird.com
    2.254.92.204.in-addr.arpa       PTR     mail.weird.com

Interesting.  I actually haven't tried this since BIND 4.  It made sense
that it wouldn't so I assumed it shouldn't and further assumed that in BIND
8 that it didn't as well.  (Sorry about that last sentence!)  Anyways, I
think you catch up with me in your next paragraph here ...

I don't remember ever having trouble with multiple PTRs in later
versions of BIND-4 either (I do remember that 4.9.7 in particular works
OK....)  I doubt I ever tried it on 4.8.3 though.....

So does the reverse resolve work correctly with the two PTR responses for
most resolvers?

I found this tidbit in my archive of the bind-workers mailing list from
back in June of 1996:

        As it stands, BIND allows an IP address to have multiple PTR
        records, but gethostbyaddr() only returns the first.

and this "reply" to a proposal to "fix" this issue:

        have you looked at the 4.9.3 or 4.9.4 version of the
        res/gethnamadr.c file, paying special attention to the
        MULTI_PTRS_ARE_ALIASES preprocessor symbol and its default
        value?

That setting is of course:

        #define MULTI_PTRS_ARE_ALIASES 1        /* XXX - experimental */

And according to my CVS repository that code's been in BIND's resolver
since before 28-Sep-94, i.e. BIND-4.9.3-BETA-9 has it but it was not in
4.9.2-940221.

I.e. Yes, most existing resolvers based on the BIND code will correctly
return multiple PTRs in responses as aliases in the returned "struct
hostent".  I have no knowledge of the behaviour of any "third party"
resolvers in this scenario though.  Obviously it's not too hard for
anyone with such a resolver, and a tool such as "host" or "nslookup"
that can be linked against that resolver, to test it though....

-- 
                                                        Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods () acm org>      <robohack!woods>
Planix, Inc. <woods () planix com>; Secrets of the Weird <woods () weird com>



Current thread: