I've already noted how I was surprised and disappointed to find that packet radio has apparently withered on the vine in amateur radio - perhaps the victim of the limited availability and convenience of transceivers with UHF bands above 70cms, or maybe other reasons why broader-band approaches can't be practically deployed in order to meaningfully increase the data rate. It looks like D-Star is available on 23cm in a solution from Icom, but I'm not sure yet if this is anything like the 'open' packet I used to enjoy tinkering with - it looks more like a closed system (to all intents and purposes), even ignoring the unfortunate use of AMBE.
My initial disappointment has been tempered however by the discovery that since I've 'been away' a system called APRS has been maturing and making excellent use of the otherwise paltry bandwidth of 1200 bps packet.
Back in the early-to-mid-nineties when I was last active, I remember that people were experimenting with internet (in its general meaning) gateways. The internet per se had yet to become ubiquitous, but some lucky folks, for instance at academic institutions, had access to the high-speed wired network over decent-speed leased lines, and local hams were experimenting. If you could construct a valid path for your connection, then even back then you could do some interesting things with some of the bridges that were in place between packet nodes around the world.
Part of the problem with AX25 'back then' was dynamic routing. You might know your destination address, but you couldn't reliably automatically route packets to it IIRC. Part of the issue here was the typical number of hops involved (each with potential issues of reliability), and the lack of a fast and also reliable means to propagate network metadata (e.g. active nodes with their lists of reachable adjacent nodes). The notion of geo-locality and therefore likely repeater paths was also missing.
While I'm still only beginning to learn about APRS, it seems that it solves most of these problems using the high-speed internet backbone for end-user message propagation and network metadata. Like other amateur amalgams of internet facilities with radio (e.g. IRLP), the internet provides a super-fast global conduit for network maintenance and administration as well as long-haul transmission, leaving the user access points using radio as the transmission medium. This reminds me a little bit of my daughter's comment "Why do you need radio Dad, when you have the internet?"... but of course she's missing certain aspects of the point ;-)
For somebody stuck in a time-warp regarding packet radio for 14 or so years, messaging across APRS seems like magic. I recently signed up with openaprs.net as a nice portal through which to begin playing with the technology, and part of the sign-up in order to use certain features requires the validation of your call sign via radio. This entails the sending of an automatically generated key back to the openaprs system. To my amazement it took me much longer to type the dozen or so characters of the key into the radio than it did to send and receive an acknowledgement over the APRS network.
What I've yet to understand about APRS is whether the infrastructure allows me to make point-to-point connections across the network - essentially tunneling a normal AX25 connection. That will be the subject of some research in due course.