Text 27538, 146 rader
Skriven 2015-09-03 12:14:36 av Michiel van der Vlist (2:280/5555)
  Kommentar till text 27436 av Kurt Weiske (46415.fidonews)
Ärende: The need for zones
==========================
Hello Kurt,

On Wednesday September 02 2015 06:54, you wrote to Alan Ianson:

 KW> I'm wondering if we need zones. Connection costs are mostly free and
 KW> zone-gating is a thing of the past.

There was never a technical need for zones and the implementation was shakey at
best. This is an article for Fidonews I wrote in April 2009. It is outdated
regarding the numbers but the essence still holds. Enjoy!


===============================================================================

                        Trimming the FidoNet hierarchy
                        By FTSC chairman, Michiel van der Vlist
                        2:2/40

Summary
-------

This is a plea to reduce the FidoNet hierarchy by at least one level. Fidonet 
has shrunk from some 36000 listed nodes in the second half of the previous 
decade to some 4500 listed node per nodelist.100 of last Friday. It is anyone's
guess how much of these are dead wood, but suspicion is that it is
considerable. This is a plea to improve the situation by reducing the number of
hierarchical levels.

Why reduce the number of levels?
--------------------------------

Soon another region will disappear from the nodelist. Possibly even by next
Friday. Region 31, Austria. It has shrunk way below critical mass and so the 
decision to not continue as a region of its own is a wise decision. Keeping up 
appearances is not compatible with "ensuring the smooth operation of the
network", which is the job of the coordinators.

This incident once again raises the question of whether FidoNet really needs
four levels of hierarchy. At ground level, there are the nodes. Above that the
net, then the regions and at the top the zones. Technically there is a fifth
level at the bottom: the points. Now that FidoNet has shrunk to a fraction of
what it ever was, do we need all those levels? Did we ever need them? Soon
after the creation of FidoNet, nets were created for two reasons; One: it
became too big for one person to maintain the nodelist. Two: to create a
default routing. I think there is no question that nets are still useful. But
do we really need two levels of hierarchy above that? I doubt it. There is no
denying that the nodelist contains a lot of dead wood and that is a sign that
the coordination hierarchy is not functioning properly, it seems to be unable
to keep the nodelist clean from dead wood. It might function better if the
hierarchy were trimmed down. Too many cooks spoil the broth.

What level to do away with?
---------------------------

Perhaps from a political point of view, it might be the best to do away with
the regions. Coming down from the top, it appears that It is at the region
level where things start to go wrong. RC's fall asleep or just disappear and
are not replaced with the result that the segment is frozen and degrades into
dead wood. However, this article is not written from the political point of
view, it is wrtten from a technical point of view and from the technival point
of view it would be a good idea do to away with the zones.

Why do away with zones?
-----------------------

For starers: we never needed them in the first place and the technical
implementation is shaky at best. When FidoNet went from 1D to 2D, all the
software was upgraded. No 1D software survived after the introduction of nets.
But when zones were introduced, Fidonet took a halfhearted turn. A system of
zonegates was introduced so that nodes could continue to use 2D only software.
Also they could operate with a nodelist of their own zone only. By now 2D only
software (stone age software) has faded out and disk space is no longer an
issue, so that "everyone" uses the full nodelist, but remnants of the 2D system
still remain. PATH and SEEN-BY in echomail are still 2D. And because net
numbers are not unique across zones, systems acting as zonegates used to strip
SEEN+BYs in interzone echomail. This causes dupe prevention not to work across
zones. As long as only ONE zonegate is used to gate echomail from one zone to
another this is no problem, but that system has long ago been abandoned.

Proposals to make PATH and SEEN-BY zone aware were filed (FSC-0052 and others)
but never implemented. As a result Circular Path Protection does not work
across zones and as it is unlikely that FSC-0052 will ever be implemented, the
only way to ever make it work is to dispense with the zones or at lest with the
interzone duplicates.

From the non-technical POV, zones have brought us nothing but grief. They have
divided Fidonet into social islands for which their was never a technical need.
Add to that, that already one zone (zone 6) was disbanded because there was no 
one left to administer it, and that two other zones are unlikely to pass
scrutiny for critical mass, there is every reason to consider continuing the
process of reducing the number of zones with the ultimate goal of doing away
with them altogether.

How to go about it?
------------------

From the technical point of view there is one problem: Nets are not unique
across zones. So when doing away with zones, we first have to deal with the
rpoblem of duplicate nets.

For starters there is the recommendation that RC's when creating new nets they
DO NOT ASSIGN NET NUMBERS THAT ARE IN USE IN ANOTHER ZONE. For ZC's the
recommendation that when RC's go against this recommendation, they do not
approve it.

How about existing nets? although the majority of them has disappeared, there
are still a few duplicate nets. All in Z1 and Z2. I counted seven. They are:

220
249
250
343
345
372
382

It seems that the problem has become manageable. Rc's may want to take a
closerlook. Some of these nets could perhaps be easily dealt with. Perhaps some
of them do not even exist any more. Others consist of only one or two sysops
that could be transferred to a neighbour net. To give just one example: net
2:220. As host it lists someone who is already listed as host in another net.
There is just one leaf node and that is a private node. If that net isn't dead,
it is in coma and should have been disbanded and have the remaining node
transferred to a neighbouring net if it is still operational. I leave it to the
reader to find other examples examples of dead or near dead nets in the
duplicate list.


Conclusion
----------

I realise that doing away with the zones will be asking a lot. It will demand
lots of diplomacy from the ZCs to come to an arrangement, if only about who
will be the new top level coordinator. Nevertheless I feel that reducing the
coordinator overhead will benefit the network in the long run and doing away
with the zones will be a good start.


=======================================================================



Cheers, Michiel

--- GoldED+/W32-MSVC 1.1.5-b20130111
 * Origin: Rheb genfu cnpvsvfg cvt fuvg chffl (2:280/5555)