[dns-operations] Signaling client protocol to authority

[Patrick W. Gilmore]

On Jan 16, 2011, at 9:05 PM, Paul Vixie wrote:

>> From: Dan Collins <en.wp.st47 at gmail.com>
>> Date: Sun, 16 Jan 2011 20:10:32 -0500
>> 
>> Forgive me, but...exactly. Isn't that the problem? Not that a DNS
>> server sends them an AAAA record that they can't use, but that they
>> asked for an AAAA record that they can't use. The problem lies in the
>> client that's making requests for an address using a protocol that it
>> ought to know it can't use, and the problem is absolutely not that the
>> server is doing exactly what the client asked it to do.
>> 
>> It isn't the job of a DNS server to decide what the DNS client
>> actually wants.
> 
> +1.

-1

It is not the job of a DNS server to do anything except what the person configuring it tells it to do.

BTW, I do agree the problem isn't the name server, it is definitely the client.  However, Grandma ain't gonna figure out her OSX is too old to do v6 right (hell, the latest one STILL does broken shit), so perhaps we can help her out in other ways?


> however, and forgive me also, this isn't about dns, this is about money.
> there is money being lost (or in some cases money not being made or made
> fast enough or often enough) on the server and cdn side because clients are
> asking for AAAA RR's they can't use.  the money is not being lost by the
> clients, nor by the ISP's of the clients, nor by the vendors of the 
> software used by the clients.  the money is being lost on the server side.
> and so, irrespective of what the clients should have done or should do,
> or what is or is not in-purview for the dns server, we're going to see a
> never-ending set of proposals from server-side vendors and operators to
> get this problem solved.

First, yes, this will probably end up being about money.  Which, to a first approximation, encompases _everything_ on the 'Net.  Anyone who says otherwise is deluding themselves.  (Or worse, trying to delude you.)

Second, it is not just server-side people who need this fixed.  You really think Comcast isn't worried about calls from broken v6 end users?  You really think if we can collectively do something to lower their call center queue that they wouldn't jump on it?  Again, anyone who thinks otherwise is deluding themselves.

Third, assume only server-side people want something.  Is that so terrible?  Suppose only client-side people want something?  Does that automatically make it bad?

Finally, I think your last four words are the most important.  "[G]et this problem solved."  It is a problem.  It does need solving.  So what exactly is wrong with proffering solutions?


As a side note, you _really_ need to stop seeing everything through the "CDN is bad" lens.  Akamai had nothing to do with my question, and I honestly don't know if anyone inside Akamai would code such a thing if there were an RFC available today.  That's not why I was asking.  (To be honest, it had nothing to do with money either, I was just musing about ways to stop the whitelisting & such.  But, since white-listing is about money, even though money wasn't my original motivation, I am guilty-as-charged - it ended up being about money in the end.)

-- 
TTFN,
patrick