Re: The state of DNS support, was Deprecating SPF

Well if we're going to risk being ridiculed, I'll go all-in. :)

At one time I had thought along the same lines - not to deprecate TXT, but rather to just say: "let's go pre-allocate a dozen new RR Types, and tell DNS servers to implement them, with undefined-text-content, and we give them random acronym names, and future IETF mechanisms can go use them on a first-come-first-served basis; and those future IETF specs can figure out a backronym expansion for the acronyms, and define what the text-content formatting is, etc."

But from reading the problems SPF listed in RFC 6686 and the emails from SPFBIS folks, I don't think that would solve the problem, even if by some miracle all DNS servers on the planet supported the new RRs.  There're just too many other DNS-related-things involved. (DNS servers and registries, and their provisioning interfaces, and DNS resolvers/caches, firewalls, CGN middleboxes, client libraries and their APIs...)

So here's the to-be-ridiculed part...
I suggest a new strategy: let the wookie win.  Instead of denying their existence, accept that there will be more folks who want to deploy new DNS-using applications quickly and without waiting for the entire Internet of DNS-related-things to change.  It's going to happen whether we want it to or not.  So we might as well try to fix whatever problems we envision it will cause, and provide guidance of how to use TXT the best way.  Right now RFC 5507 only warns folks against it, but I don't see a "if you are going to use TXT anyway, then do it this way" type thing.  Nor do I see an IANA registry of prefix names, nor an RFC describing a general solution to the wildcarding problem when prefixing.  I think those things are possible, maybe.

We can require abstinence or we can give out condoms.  Which one do you think is more likely to prevent spread of disease?

-hadriel


On Aug 25, 2013, at 11:36 PM, "Dickson, Brian" <[email protected]> wrote:

> At the risk of ridicule, and in the interest of potentially useful results:
> 
> Why not deprecate TXT, or at least abandon the current TXT and introduce
> a new TXT code point? And, as each iteration gets trashed, move on to the
> next one?
> 
> It's not like there is a shortage of type code values, and it is not like
> the RDATA of TXT is complicated, or in any way is required to be unique.
> SPF is, after all, exactly the same as TXT except for being different.
> 
> If SPF want to take over TXT, let them. We can all move to TXT2, until
> someone ruins it, and then we can introduce TXT3, etc.
> 
> I realize it is rather silly, but if it works, and it disambiguates things,
> the fact that it _is_ silly is perhaps not the greatest reason for not
> doing it.
> 
> This might provide a small clue to the next group of other-protocol folk,
> that using TXT instead of asking for their own RRTYPE, is foolish, unwise,
> and likely to result in having a new RRTYPE lobbed at them in retaliation.
> :-)
> 
> (My two cents, late to the discussion.)
> 
> Brian
> 

_______________________________________________
dnsext mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsext