Reality Check

Wed Mar 14 21:13:17 UTC 2001

On Wed, 14 Mar 2001, Timothy R. McKee wrote:

> > Bear in mind that in many cases, this is an illusion.  They aren't
> > accessing the same machine at all.  Someone is using round robin DNS
> > to map one name into several IP addresses, or a Local Director to
> > map one IP address into many IP addresses, or there is some other such
> > substitution being employed.
> >
> > In some cases the party serving the data is involved in the illusion.
> > In others, as in transparent proxying, someone along the way is
> > intervening.  This is often silent and may have the consent of neither
> > the user/client or whoever is running the intended target.
> 
> Yet in all cases, except where something is physically broken or out of
> synch, the initiating user and the terminating server expect that access to
> information or services via a documented public mnemonic URL will provide
> the same information (or a cached copy of it) to every user globally.  If it
> doesn't WE are the ones that are held responsible by the users.

That may be true.  Nevertheless, the existing system works.  Whenever
it fails to work, we fix it.

> > My point is that we are already in the world that you are warning us
> > about.  People are happily using one address space within their
> > company and quite another to talk to the outside world, with NAT
> > mediating between the two.  Their internal DNS is also different from
> > the DNS seen on the global Internet.  And it all seems to be working
> > exceedingly well, despite the fact the games people play with IP
> > addresses and domain names are becoming very subtle indeed.
> 
> But once again, when they access or publish a PUBLIC URL, they have
> expectations  that it will work and it will work the same for everyone
> regardless of location or ISP affiliation.  I don't consider internal
> network workings to be public in nature.

And that is of course true.  But we still have a working system, 
one in which we, the network operators, ensure that our customers
are content that the illusion with which we present them is correct.

Now consider the relative complexity of the systems involved.  In
one, there are at least tens of thousands of address fiddles in
operation, cases in which the machine you think you are accessing
is not the machine that you are really accessing.   This goes on 
all the time.  Moderately often there are glitches.  When there 
are, we fix it or complain to someone who can.  Generally speaking,
problems get resolved fairly quickly.

In the other case, there are a couple of hundred mappings from 
domain names like .COM, .UK, .NET, .FR into the IP addresses of name
servers authoritative for these names.  Someone else pointed out 
that the information involve is around 60 KB in all.  Please 
convince me that the world's ISPs are not capable of managing 
this simple task.

Spelling out the obvious: let's say that VBCnet started referring 
our customers to the wrong name server to resolve names in .COM.
How many minutes would it be before the phones began ringing off 
the hook?  I can assure you that we would fix it really fast, and
take steps to make sure that we didn't screw up again.

--
Jim Dixon                  VBCnet GB Ltd           http://www.vbc.net
tel +44 117 929 1316                             fax +44 117 927 2015