Text 11266, 146 rader
Skriven 2005-07-27 14:08:00 av Michiel van der Vlist (2:280/5555)
  Kommentar till text 11262 av Matt Bedynek (1:106/1)
Ärende: Fidonews Echo Rules
===========================
Hello Matt,

 >>> I do not beleive that will solve anything at all.

 MvdV>> Well, you have to keep in mind that because we have an echopol, EC's
 MvdV>> actually have some authority. Or had, when out of area feeds were
 MvdV>> prohobitively expensive. They had the authority to order link cuts.

 > First of all, like the rest of Fidonet these days, your ECs do
 > not operate exactly how ECHOPOL defines them to operate.

I know that some interpret it that way. It does not matter. The ones who work
with it se it different and that is all that matters.

 > Much like policy is used in this zone, only parts that are
 > convenient are used.  That does not mean the practice is bad
 > but that perhaps the document needs revision to define current
 > practice.

There is as much chance of revising EP1 as there is of revising P4. Several
attempts were made, all have failed.

 MvdV>> Yes, that is a problem. Over here the fully connected polygon was
 MvdV>> never very popular. Maybe for the reason you point out. The most
 MvdV>> complex I ever saw in my own region was a simple triangle.

 > My opinion is that it is simply not needed.

I agree. There is no added value in complex polygons. Not today anyway. The
speed advantage is negligible and the SPOF /still/ remains.

 > Everybody wants to be a "big mail hub" but not everyone is cut
 > out for the job.

Indeed....

 > Broadband made becoming a mail distributor too easy.

It has made it affordable, not easier...

 > Not that I have a problem with this but there is more to
 > hubbing mail than setting up a tosser and linking other nodes.
 > When you start forwarding mail, you are responsible for what
 > ever your system injects into the stream.  And, most of the
 > time you are alone in isolating problems and soliciting
 > cooperation to get them fixed.  When someone that hubs fails
 > to do their job they are passing the buck to their peers.
 > That is the current state.

 > Topology grows faster than the network has a need for.

The network is not growing, on the contrary it is shrinking. If we did things
right, the topology hpulkd become less complex. It seems we are doing something
worng...

 > Fidonet is not so mission critical that large meshing is
 > needed.

There never was a need IMHO. There could have been an advantage if we had
introcude flooding techniques as was done in newsnet. A group of German sysops
tried to develop something like that. It failed because it hinged around unique
MSID's and they could not convince the FidoNet community to make MSID's
compulsory.

 > Despite the freedom that broadband has given people.  A
 > coordinated structure (with some technical sanity) runs much
 > better.  Fidonet, as it exists, is a free-for-all.  That is
 > why it needed routing charts (which you were lucky to have
 > updated - if at all).

We never had routing charts in Z2. Not on a zone wide basis anyway.

 > I just remember the technical ceiling of the average hub 10
 > yrs ago being much higher than it is today.

Not just the hubs. Also the average technical competence of the sysops has
dropped markedly.

 > It was cool to see a routed netmail reach the destination in
 > a timely manner even going through a dozen hops!  Today, you
 > are lucky to receive routed netmail even if it has to cross
 > only a few.

Indeed. That is why I try to dliver netmail as close to the destination as
practical these days. To increase the chances of it getting delivered. Routed
netmail is not as reliable as it used to be. Not by far. :-(

 > Not only that, the network was so much larger.  I'd hate to
 > see the state of things in this zone today if the same number
 > of nodes were present.

It would be a mess. Lets' face it, the majority of the technically competent
has left and that comptency is lost forever.

 > Well, I'd like to see the nodes but not the mail problems.  :-)

In your dreams... ;-)

 MvdV>> Yep. And maybe that is why we normally saw just star like distribution
 MvdV>> here. Complex fully connected polygons are difficult to coordinate.

 > That is why we have the current configuration.  Some wanted a
 > five system mesh but I was strongly against it.

And for good reasons I think. The number of links is (almost) proportional with
the square of the number of stars. That means you (almost) double the potential
problems if you go from four to five stars.

 > Instead, I felt satellite meshes composing of no more than
 > three systems was best.  And, there were already mesh
 > clouds operating so why not capitalize on that!

With today's resources one hub can easily feed 50 downlinks. So with four
toplevel hubs each feeding 50 endlevel hubs that in turn each feed 50 end
nodels in a star configuration, we can serve 10.000 nodes. And node node would
be more than five hops away from any other node.

And of course the sensible thing to do would be to make these endlevel hubs the
host of the net that consists of the endnodes they feed. And the four top level
hubs would be the four NCs of a four region FidoNet.

That's all we really need today.

Yeah, I know... Don't wake me up yet, let me enjoy the dream just a little
longer. ;-)

 MvdV>> Also note that over here emphasis was always more on dupe prevention
 MvdV>> rather than dupe detection. In a star like configuration it is fairly
 MvdV>> easy to see what link shoud be cut to break the dupe loop and as I
 MvdV>> said, EC's had the authority to order kink cuts.

 MvdV>> It worked fairly well.

 > You cannot prevent what you cannot control.

One can control it if the configuration is not needlessly complex and one has a
set of rules that everybody follows. (Voluntary or not). As was proven in R28
during the heydays of Fidonet.

Not that it was any better than the anarchism of today, but it proves that it
can be done.

Cheers, Michiel

---
 * Origin: http://www.vlist.nodelist.org (2:280/5555)