New subject: My view to 44net's future

11 Jul 2020

It is about a year ago that I tried to discuss such a proposal.

My view was like this: let's establish routers at datacenters around the 
world, in
addition to the existing UCSD router and some others that already handle 
/16 networks
on internet.   I was thinking about around 10 routers globally. They 
interconnect using
BGP on private AS numbers (the 32-bit AS numbering scheme we already 
use) over a mesh
of tunnels between them, to exchange routing information for 44Net 
subnets, but on
internet the announcing remains as it is (i.e. the whole network is 
announced at UCSD,
and regional subnets are, but do not need to be, announced at those 
global routers).

The "users" connect to those routers using a (small) variety of tunnel 
protocols to match
the restrictions they face from their internet providers, e.g. forcibly 
being behind a NAT
router, having a dynamic IP address, maybe having some enforced 
firewalling, etc.
I was thinking of standard tunneling protocols like GRE, GRE/IPsec, 
L2TP/IPsec, etc.

The tunnels would be operated in a point-to-point fashion by default 
(/30 or /31 subnets
on the tunnel), and the users would use BGP to announce their own 
routable subnet over that.
They can setup redundant tunnels to multiple global routers when they 
desire to do so.
They can also setup tunnels directly to other users when desired, and 
run a BGP session
with them.  And of course, radio links can be incorporated in the scheme.

Users could use the widely available inexpensive routers available today 
that can use
these standard protocols without special tinkering with scripting, 
locally compiled
executables, etc.   E.g. the inexpensive models available from MikroTik, 
Ubiquiti, etc.
More technically inclined users could install software on their own 
Linux system or -board.

I see this as a solution for the following problems:
- more and more users struggle with getting IPIP routed on their 
internet connection, due
to them being behind ISP-managed routers, CGNAT, having dynamic 
addresses, etc.
- non-technical users struggle to get our special IPIP mesh operational 
on their routers,
where using industry-standard protocols would be much easier as their 
router config
interface already knows about those.
- some users requested to have redundant IPIP tunnels (multiple internet 
routers serving
the same 44Net subnet(s) in a redundant way, which the IPIP mesh cannot do.
- the IPIP mesh does not really allow to check the status of the 
configured gateway
routers, so routers that have not been operational for a long time just 
remain in the tables.

I expected enthousiasm from the users, but unfortunately I was met with 
a lot of
resistance to change, e.g. from people who believed that such a system 
would rob them
from their direct tunnel to their buddies on the other side of the world 
and force them
to go via established and centrally managed hubs (incorrect, of 
course).  As I see this as
a hobby and not as a struggle to be right and convince those that do not 
want to be
convinced, I did not pursue it further.

I don't know if the attitude an scepticism has gone away now.  We would 
have to see in
a new discussion.  Maybe some of the opponents have realized by now that 
it would be
better to have a more flexible mechanism like this instead of going on 
with the IPIP mesh
forever.  Maybe not.

I don's see the need of routing the entire 44Net from internet to all 
those routers.  When
everyone routes only their own regional subnet(s), it remains more 
manageble and we
do not face the raised issues.  Furthermore, some of us have our ISP 
announce the
relevant regional subnet on their redundant border routers under their 
AS, and then
route it to our "gateway" router.  That relieves us from being 
responsible for that
announcement, and we use the ISP NOC services.  But of course it also 
means we are
less integrated with the internet routing, e.g. we cannot allow that 
subnets from our
announcement are routed to others.
Of course everyone can decide if they want to announce their subnet on 
internet
directly or via an ISP, but I would suggest that the internet side of 
things be kept separate
from our internal routing (2 BGP instances, the 44Net one using a 
private AS number)

W.r.t. the .ham TLD: I don't see what advantage that would bring, we 
already have the
.ampr.org domain and we run the DNS for it.  It should offer the same 
capabilities as
a dedicated TLD, I think, at a much lower cost.

Rob

Re: [44net] My view to 44net's future