[Top] [All Lists]

Re: Impending publication: draft-iab-dns-choices-05

2008-03-04 06:53:42

On Fri, Feb 29, 2008 at 09:57:15AM -0800,
 Olaf Kolkman <olaf(_at_)nlnetlabs(_dot_)nl> wrote 
 a message of 48 lines which said:

The IAB is ready to ask the RFC-Editor to publish

    Design Choices When Expanding DNS

I fully agree with the overall message ("the use of a new DNS Resource
Record Type is the best solution") but not with the way that the other
choices are ridiculed:

DNS extension discussions too often focus on reuse of the TXT
Resource Record Type.

It is a bit strange that there is nowhere in the draft even a small
bit of auto-critique. If many people used TXT RR types, it is because
it has been historically quite difficult to get a new RR type
registered (the problem which is addressed by RFC 2929bis).

        It was historically difficult because:

        * there wasn't a standard mechanism to emit a unknown record
          in master file format.
        * the description of when to use compression pointers was
        * some nameserver only handled known type (through design or
          as a result of implementation bugs).
        * there was no clear proceedure for gettting a new RR type

        All of these issues have been addressed for years now.
A second reason why many groups used TXT RR is because middleboxes

        If you worry too much about middle boxes you will never
        deploy anything new.  People will replace middle boxes that
        interfere with their need.

and at least one Microsoft API do not allow unknown RR types. This has
been reported several times.

        So you want to hold up everything because one company
        produced a API that was not RFC 1123 compliant?

Reading the current I-D, one may wonder why so many people stupidly
used TXT records... (See also the very questionable "it is worth
reexamining the oft-jumped-to conclusion that specifying a new
Resource Record Type is hard" in the conclusion.) 

        Because people keep putting up FUD whenever anyone suggests
        getting their own RR type.

        Having got a new RR type, it really was not that hard and
        that was a politically sensative RR.  Others have gone
        straight through once there was agreement about what to put
        in them.  You have that problem whether you use a TXT record
        or not.

        Also you don't have to use TXT to experiment.  There are
        enough types available to be able to experiment provided
        you actually go to the effort of getting a proper RR when
        you are done.  Been there, done that too.  About to do it

example of new data is [...] data used for fighting spam 

If this refers to RFC 4408 or 4871, then "fighting spam" is not a
correct summary (by itself, authentifying the sender does not fight
spam, spammers can be authentified too). "Email domain sender
authentication" may be a better term.

Splitting an RRSet is a protocol violation

I have a doubt here. The protocol allows to send partial RRSets, you
just have to set the TC bit (RFC 2181, 5.1).

The process for creating a new Resource Record Type is specified in

It is just an informative reference. Do you plan to publish
draft-iab-dns-choices-05 before 2929bis? (It would be a bad idea,
IMHO, to tell people they must use a new RR type, before the new
procedure for doing so is ready.)
IETF mailing list
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark_Andrews(_at_)isc(_dot_)org
IETF mailing list