John Yes, but my thoughts are why use a relational database to do the job of a Routing Protocol ?
Something like ISIS or OSPF is easily able to cope with the number of repeaters and stations we have. And having the Private IP Address space is actually pretty good as we can keep it to an IP VPN. I'd only be making a look-up between the callsign and an IP address. The actual payload be-it UDP or TCP is pretty irrelevant; you just need to know how to get from A to B which today is source-routed on G2. DNS is scalable to the extent that we need as well (probably 4 servers worldwide), plus the local machine could be a caching nameserver taking strain off the central ones (which wouldn't actually be a strain at all given the low numbers even if every Amateur in the world had D-Star). I know of similar SQL-based routing systems that struggle to survive at the 10,000+ user level (I'll let you guess...) but IP routing at that sort of figure is a walk-in-the-park. If it was done properly, we could even get G2 callsign routing on DV-Dongles. I'm going to put some ideas forward to the Trust Server team and see what they think, but there certainly is a lot of scope for getting some real experimentation back into Amateur Radio. That alone is a very good thing in these days of black-box radios IMHO. 73 David - G4ULF --- In dstar_digital@yahoogroups.com, John Hays <j...@...> wrote: > > David, > > I'm only Cisco IOS literate with a manual in hand :) but I see the > appeal of the reuse of IP protocols. I just think it confuses the > issue in D-STAR > > I feel the callsign (including the 8th character designator) in the D- > STAR protocol is unique enough for endpoint determination. > > The routing is pretty simple in my "simple" mind. > > Gateway receives a D-STAR header. > If it knows the destination (URCALL) already (say its on a locally > attached repeater or at a cached gateway location), it sends it to the > destination. > If it doesn't know the destination it asks the "trust server" where > to deliver it (probably the IP of the gateway to last report seeing > the callsign) > If it can't find the destination, it sends a service message back to > the origin callsign that the destination callsign is unknown. > > The gateway also looks at the source (MYCALL) > If it is already cached as being on an attached repeater, do > nothing > If it is from an attached repeater and either doesn't exist or > is > marked being on another gateway, update the trust that the callsign is > now on this gateway > If it is from a remote gateway, update local cache to associate > the > callsign with the remote gateway > > Of course there is more code to it than that, to handle routing > loops, timeouts, etc., but I think it pretty much handles > "connectionless" D-STAR traffic. > > For DD, this means you can run any Ethernet encapsulated protocol > between endpoints, not just TCP/IP. You can also use the address > space that makes sense, e.g. 10.x.x.x, 44.x.x.x, 192.168.x.x, > 172.x.x.x, or x.x.x.x in the IP world. > > John > > On Jul 24, 2009, at 1:40 PM, dlake02 wrote: > > > John > > > > The whole use of the 10.X.X.X addresses seems like a real mess, but > > in fact, it could be put to very good use. > > > > The advantage is that every callsign has a unique address in the > > system, down to a device level. > > > > A very similar concept exist in the cellular networks - each device > > has it's own identity and that identity moves between cellular zones > > and networks. Whilst callsigns are useful, there is no good routing > > protocol for callsigns - that is what G2 attempts and fails to do > > today. > > > > Now, host addresses actually are very useful things, and there is no > > reason why these couldn't be used to provide a much more scalable G2 > > architecture that retained it's compatibility with the existing G2 > > network. Advertising the movement of a /32 address across a small IP > > network such as the G2 network using an IETF standard routing > > protocol would be very quick. > > > > I'm not going to go into great detail on my thoughts - there are > > correct forums for doing that, and that isn't here. However, for > > those that are Cisco IOS literate think about this: > > > > - Three routers, connected via your favourite IGRP (mine is OSPF). > > - DNS with a lookup to G4ULF 10.1.1.1 > > - OSPF on both routers with "redistribute connected" configured > > - Create a loopback on R1 for 10.1.1.1/32. From R3 ping "g4ulf" > > - Delete the loopback on R1 and create on R2 for 10.1.1.1/32. From > > R3 ping "g4ulf" > > > > Now, pretend the routers are G2 gateways and the loopback is created > > from the "Last Heard" table.... > > > > David - G4ULF > > > > > > > > > > John Hays > Amateur Radio: K7VE > j...@... >
