[6bone] Request: two 6bone pTLAs

[Jørgen Hovland]

----- Original Message ----- 
From: "Paul Jakma" <paul at clubi.ie>
To: "Jørgen Hovland" <jorgen at hovland.cx>
Cc: <6bone at mailman.isi.edu>; "Lars-Johan Liman" <liman at autonomica.se>
Sent: Wednesday, May 12, 2004 6:36 PM
Subject: Re: [6bone] Request: two 6bone pTLAs


> On Mon, 10 May 2004, Jørgen Hovland wrote:
>
> RA does one thing, does it well. Resolver server addresses, etc, are
> outside its scope.

Yes, and that's the problem with it.

>
> > To the contrary, I find this very convincing.  DNS is an important
> > service on the internet - but it is certainly not mandatory for
> > every single networked system on earth.
>
> I dont think Iljitsch is advocating that DNS be mandatory for all
> systems. I think he's asking to expirement with having a well-known
> address for recursive DNS service, so that hosts which _want_ to
> resolve addresses can use (by prior knowledge) an address which is
> guaranteed to work, while obviously still retaining ability to
> manually configure in resolver addresses.
>
> Imaginary configurability of implementation issues should have no
> bearing on whether a WKA for recursive DNS service is a good idea or
> not.
>
> > I must say that I agree with Mr Liman on this. DHCP works today and I
don't
> > see why the concept shouldn't work tomorrow.
>
> Not everyone wants to run a DHCP client.
>
> Directing traffic for a WKA address to a suitable DNS server(s) would
> be far easier.

You still need to actually _know_ if the network use WKA or not. Determing
that manually or by probing is not an acceptable solution. If you want WKA
to work automaticly, you would need a way to advertise on the network that
"we use WKA here", like RA. Doing that for all services cause N times amount
of noise which could perhaps be done better dhcp style. It would be better
to use something that could advertise the whole package (ntp, proxy, wins,
bootp, nntp, smtp etc) maybe even depending on who you are (eg mac).
Something like dhcp. Well, at least that's what I think.


>
> I support Iljitsch's idea to try this as an experiment.
>
> I disagree that it requires two pTLAs. My feeling is that there
> should absolutely _NOT_ be any public DNS recursive service offered
> at the WKA because of the security implications of a widely used
> public recursive DNS service. Even as an experimental address, it
> should not be public, because of the risk of it becoming widely used.
>
> The WKA should be confined internally to organisations, as a
> convenience, should they wish to make use of it.

Seconded.


Joergen Hovland