Setting up a Root name server

[Scott Morizot]

On 6 Sep 99, at 22:23, Barry Margolin wrote:
> In article <37D40440.57406722 at megabytecoffee.com>,
> chris  <chris at megabytecoffee.com> wrote:
> >I only disagree when experts when they tell me that something can't be done that
> >I'm sure can be.
> 
> No one said it can't be done, and it's pretty obvious that it *will* speed
> things up (not recursing has to be faster than recursing), assuming you put
> enough RAM in your server.  What we're saying is that the speedup won't be
> as enormous as you seem to expect, so it's not worth doing.  The DNS
> caching architecture was specifically designed to make such a configuration
> unnecessary -- your server will usually have all the delegation records
> that it needs.

Actually, I sent a detailed response outlining why the action he
proposed would not produce the result he clearly wanted.  He believes
that a name server uses the hints file as an authoritative source
of root name servers.  It does not.  It uses it to find them.
Once it finds one that responds, it replaces the non-authoritative
information from the hints file with the authoritative list
of name servers returned in the response.  So a name server with his
modified hints file has, at best, one-thirteenth chance of hitting 
his fake root on startup.  If it doesn't, it will never see it or 
query it at all.

I went into more detail in my earlier response since I have some
insight in the matter.  I run one of the internal roots at work.
But he apparently ignored that response.

After my response, I saw one from Cricket that said essentially
the same thing in much less space than mine.  He ignored
that one too.

But he has had people tell him that what he proposes won't
work as he obviously wishes.  And it's been explained to him
why.  If he chooses to do so anyway, let him.  It won't impact
the rest of us.

Scott