Further thoughts on VUNIC [was Re: Announcing VUNIC/VUnet]

From: Brian Chase <bdc_at_world.std.com>
Date: Sat Jun 16 12:02:54 2001

On Fri, 15 Jun 2001, Eric Dittman wrote:

> > > > Can you kindly reserve the name "mystica" in your UUCP map
> > >
> > > I'd like to keep using "kumiss"
> >
> > Please give me "lepton". I've never even *used* UUCP before, but
> > darnit, this will be fun and educational.
>
> I'd like "narnia", which is my old UUCP node name.

It sounds like VUNIC/VUnet will need to implement a central name
registration service as one of its first tasks :-)

Might I suggest that as part of this name registration database, people
also record their ICBM coordinates. This way, not only can we define a
logical structure for the UUCP network, but we can project it onto a
physical map as well.

Another big question which arises is what will the topology of the network
be like? At least as far as TCP/IP connected nodes are concerned. Back in
the original days of UUCP, this was significantly influenced by the costs
of local and toll phone charges. You'd want to minimize the phone service
costs by connecting to the least expensive site for your UUCP feed.

With TCP/IP internet access, it now becomes equally cheap to connect to
any other TCP/IP based VUNIC site. So which path do you choose? Or does
that even matter? I first got onto the net in 1989. UUCP was still in
wide use, but by the time I got fairly sophisticated in my net knowledge,
TCP/IP had almost completely displaced UUCP as a means for pushing data
around.

I do thing it would be very smart, and perhaps immensely useful later,
to register node info in some sort of central repository with replication
abilities. I would strongly suggest something LDAP based, like say the
UMich OpenLDAP server. And then for older systems without support for the
LDAP client API, we could perhaps design some lighter-weight means by
which they can access the same information stored in an LDAP server.

Types of information which I think would be useful to record for each node
would include things like the following:

  * Unique site name
  * Unique site ID#
  * ID(s) of sites upstream to which you can connect.
  * Latitude/longitude coordinates
  * Whether the node supports TCP/IP connectivity and at what speed.
  * Dialup phone number(s) by which the site is accessible.
    (Maybe broken down into subfields which include country access code,
     region code, and then the phone number).
  * (Packet radio specific information?)

With some of this information, it should be possible to dynamically
structure the maps in such a way that the highest speed TCP/IP based sites
can form a "backbone" for traffic.

Okay, so it's maybe overkill for how our traffic would be taxing things :-)
Hey, isn't UUCP considered peer-to-peer networking?

-brian.
Received on Sat Jun 16 2001 - 12:02:54 BST