On 11/8/21 6:00 pm, Toussaint OTTAVI via 44Net wrote:
Our current design is the result of several iterations in the last few years, and from several talks with Jann DG8NGN, (one of the founders and network managers of European Hamnet, and currently Chairman of the TAC) :
- Our first design was using 10.44.0.0/16 "private" addressing and
NAT. Sufficient for inter-connecting sites, but NAT headaches for inbound traffic.
- Then, we migrated to 44.168 "Intranet" addressing, Hamnet-style, but
with no IP-IP implementation (dislike the tech, and no current need to route outside of the island)
- When 44.190 specs were published for Internet-connected things, we
got a subnet and announced it using BGP and a $5 Vultr VPS. We deployed it for our XLX, D-Star and DMR stuff.
- As we were testing, and we did not know in advance which addressing
scheme would be the best for a specific situation, we decided to implement dual addressing on all locations.
44.190 came about right at the time I was about to renumber, and has been a great asset for me.
Dual-addressing is a bit tricky to setup when you start from scratch. But once understood and implemented on a model with a "POP" and "Access routers", it's just a matter of copy-paste of configurations. As we are on a very small scale, we do it manually. But there should be no problems to write configuration-builder scripts for use on a larger scale.
I have done similar before. I run several networks on the one wire, so I'm no stranger to complex routing setups. :)
I've got over 200 IPs, and the most annoying part is that a couple of the addresses require manual intervention by the administrators of services (D-STAR REF admin and IRLP admins),
This would not be necessary if those admins/designers used DNS names instead of static IPs in their systems, HI :-) I never understood why FQDN names are not used more in D-Star / DMR / Digital modes in general...
IRLP actually uses its own name resolution, which works well for IRLP nodws, even those on dynamic IPs. But reflectors are assumed to be on static IPs and appear to be hardcoded into the system. And it seems something similar for D-STAR REFxxx reflectors. But I use FQDNs wherever possible, for the reasons you state. And that also makes enabling IPv6, when it becomes available for a service easy (Just add an AAAA record and stir gently ;) ).
However. I am open to this complex renumbering operation, if the proposal results in a viable long term structure. I can certainly see the routing advantages, the more I think about it. So, I'm conditionally for this proposal - the condition is that it's done properly. I am one who wants both Internet and Intranet connectivity.
+1. I am a network designer. I don't want to choose or force something for my users. Every end-user should be able to choose which addressing scheme to use according to its own needs. As network admins, I think we have to provide both addressing schemes to our users, in a simpler and more standardized way.
I see this as potentially simplifying routing for those of us running dual ported systems.