In article <1181066741.622086.276610@p77g2000hsh.googlegroups. com>,
wrote:
>Is it possible to configure a DNS resolver to start a domain name
>resolution with quering top-level domain nameservers without sending
>any queries to root nameservers?
>I'm not a DNS server administrator, just interested how easy is it to
>implement. Downloading root.zone file which size is 68 KB (instead of
>named.root) is not a problem anymore. Thus I'm asking myself: Why
>resolving root domain if one can start directly with TLD?

Why stop there, why not have a file with the most popular hostnames so
you don't have to track down www.google.com?

The simple answer is DNS was designed to be able to start with just a
few hard coded (named.root) values.

The root nameservers are very rarely changed, the last time that
named.root changed was 2004-01-29, the last time root.zone changed was
2007-06-04. Thats yesterday by my calendar.

--
-- Rod --
rodd(at)polylogics(dot)com

In article <1181066741.622086.276610@p77g2000hsh.googlegroups. com>,
silentser@gmail.com wrote:

> Hi!
> Is it possible to configure a DNS resolver to start a domain name
> resolution with quering top-level domain nameservers without sending
> any queries to root nameservers?
> I'm not a DNS server administrator, just interested how easy is it to
> implement. Downloading root.zone file which size is 68 KB (instead of
> named.root) is not a problem anymore. Thus I'm asking myself: Why
> resolving root domain if one can start directly with TLD?

Download the root.zone file and configure your server as a master for
the "." zone with this as the source file.

--
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***
*** PLEASE don't copy me on replies, I'll read them in the group ***

On Jun 5, 8:40 pm, r...@panix.com (Rod Dorman) wrote:
> In article <1181066741.622086.276...@p77g2000hsh.googlegroups. com>,
>
> wrote:
> >Is it possible to configure a DNS resolver to start a domain name
> >resolution with quering top-level domain nameservers without sending
> >any queries to root nameservers?
> >I'm not a DNS server administrator, just interested how easy is it to
> >implement. Downloading root.zone file which size is 68 KB (instead of
> >named.root) is not a problem anymore. Thus I'm asking myself: Why
> >resolving root domain if one can start directly with TLD?
>
> Why stop there, why not have a file with the most popular hostnames so
> you don't have to track downwww.google.com?
>
> The simple answer is DNS was designed to be able to start with just a
> few hard coded (named.root) values.
>
> The root nameservers are very rarely changed, the last time that
> named.root changed was 2004-01-29, the last time root.zone changed was
> 2007-06-04. Thats yesterday by my calendar.
>
> --
> -- Rod --
> rodd(at)polylogics(dot)com

Ok, that's the way it was designed almost 20 years ago. Still,
root.zone file is changed not so often (maybe just a couple of times
per week). Synchronizing it between DNS resolvers of ISPs and large
organizations running their own DNS servers is quite feasible now.
And, probably, even easier than running a bunch of root nameservers.

Sergei

On Jun 6, 1:38 am, silent...@gmail.com wrote:
> Ok, that's the way it was designed almost 20 years ago. Still,
> root.zone file is changed not so often (maybe just a couple of times
> per week). Synchronizing it between DNS resolvers of ISPs and large
> organizations running their own DNS servers is quite feasible now.
> And, probably, even easier than running a bunch of root nameservers.


There are a huge number of name servers running in the world, you'd
have to reconfigure all of those to either download the root zone file
from someplace or query some other DNS (like you ISPs) instead.

And what would be the advantage? If you have to set up a bunch of
redundant servers to allow people to download the TLD zone file, what
have you gained? By definition, that's what the root servers already
do (except one entry at a time), without a secondary mechanism. And
mind you that every name server in the world (excepting some
misconfigured ones) caches that stuff anyway, and most (all?) of the
TLD entries have TTLs of a couple of days, so you're local name server
would need to look up ".com" only once every other day anyway.

Nor is the load on the root servers particularly high. Since everyone
caches those results, and they have long TTLs, it's rare that any
given name server ever recurses all the way to the root servers. In
fact, most of the queries that the root servers actually process are
errors (parts of malformed domain names, etc.) or unnecessary, numbers
up to 98% have been reported. There are recommendations for people
who maintain name servers for things they can do to reduce that
unnecessary traffic (see RFC4697, for example), none of which are
disruptive.

The root servers themselves are fairly modest machines too. In terms
of load, statistics are available on a number of them at:

http://www.root-servers.org/

"H" for example, is a cluster of three machines and averages around 9K
queries per second (for the whole cluster). "K" is a distributed name
server with five main mirror sites and a dozen minor ones (IP anycast
is used to distribute requests), and the total for all five main
mirrors is about 10K queries/second, ranging from 6K+ in London to
about 300 in Miami. The others have similar loads. I'm not sure you
could actually buy a desktop PC that couldn't handle the load.

So the question exactly what problem are you trying to solve by adding
this significant complication to the world? And if you really want
to, just run your own root server and download the root file from
http://www.internic.net/zones/root.zone on a regular basis.

On Jun 7, 1:49 am, "robertwess...@yahoo.com"
wrote:
> On Jun 6, 1:38 am, silent...@gmail.com wrote:
>
> > Ok, that's the way it was designed almost 20 years ago. Still,
> > root.zone file is changed not so often (maybe just a couple of times
> > per week). Synchronizing it between DNS resolvers of ISPs and large
> > organizations running their own DNS servers is quite feasible now.
> > And, probably, even easier than running a bunch of root nameservers.
>
> There are a huge number of name servers running in the world, you'd
> have to reconfigure all of those to either download the root zone file
> from someplace or query some other DNS (like you ISPs) instead.
>
> And what would be the advantage? If you have to set up a bunch of
> redundant servers to allow people to download the TLD zone file, what
> have you gained? By definition, that's what the root servers already
> do (except one entry at a time), without a secondary mechanism. And
> mind you that every name server in the world (excepting some
> misconfigured ones) caches that stuff anyway, and most (all?) of the
> TLD entries have TTLs of a couple of days, so you're local name server
> would need to look up ".com" only once every other day anyway.
>
> Nor is the load on the root servers particularly high. Since everyone
> caches those results, and they have long TTLs, it's rare that any
> given name server ever recurses all the way to the root servers. In
> fact, most of the queries that the root servers actually process are
> errors (parts of malformed domain names, etc.) or unnecessary, numbers
> up to 98% have been reported. There are recommendations for people
> who maintain name servers for things they can do to reduce that
> unnecessary traffic (see RFC4697, for example), none of which are
> disruptive.
>
> The root servers themselves are fairly modest machines too. In terms
> of load, statistics are available on a number of them at:
>
> http://www.root-servers.org/
>
> "H" for example, is a cluster of three machines and averages around 9K
> queries per second (for the whole cluster). "K" is a distributed name
> server with five main mirror sites and a dozen minor ones (IP anycast
> is used to distribute requests), and the total for all five main
> mirrors is about 10K queries/second, ranging from 6K+ in London to
> about 300 in Miami. The others have similar loads. I'm not sure you
> could actually buy a desktop PC that couldn't handle the load.
>
> So the question exactly what problem are you trying to solve by adding
> this significant complication to the world? And if you really want
> to, just run your own root server and download the root file fromhttp://www.internic.net/zones/root.zoneon a regular basis.

Thank you!

Sergei