I wonder how that is compares to AirTalk
The short answer here is 'the speed of a network is a much more complicated thing than you seem to think'.
The long answer is: these are totally different beasts. The Farallon box just takes LocalTalk, which is 230.4 kbps, and modulates that over IR, then back again. No packet-level processing is done. It's just a media converter, so it can operate at very low latency, but it is literally just like a wireless wire with all its benefits and shortcomings.
But note that the 230.4 kbps of wired LocalTalk, and of the Farallon box, is
line speed, and line speed does not mean actually achievable useful throughput, for a very large number of reasons. Line speed is the speed with which an individual bit or byte is transmitted down the line, it's the clock rate of the line itself.
First of all, you have the framing overhead, the packet bureaucracy, the mucking about that is necessary to get frames and packets to their intended destinations. You have collision avoidance. You have the fact that the machines are slow (one could, in theory, wedge a 10gbit Ethernet interface onto a 68000. Think it could push 10gig? Nope). And then you have latency, and latency is the big one.
Latency is the amount of time that a packet takes 'in flight' between two network nodes. It's the time it takes between leaving the source node and arriving at the destination node. And the problem with it is that during that time, both the source and destination are just sitting there scratching themselves and not doing much useful. And adding 'wait time' to the middle of a transmission means that the transmission as a whole takes longer, even if each individual packet in flight is being transmitted at line speed.
The link between AirTalk and the Mac is 230.4 kbit/sec. It has to be, or the whole thing wouldn't work, because that's the defined line rate of LocalTalk. The WiFi segment at the other end is running a lot faster. And yet the whole system has, basically unavoidably, less throughput
for any given two machines, because the latency of IP multicast over WiFi is wildly unpredictable compared to the very strict latency requirements for wired LocalTalk. When you add Mini vMac and desktop OS's packet scheduling into the mix, it becomes more so, and this is fundamenta,ly why speeds into mini vMac tend to be a bit questionable over WiFi.
But even this isn't the full story. AirTalk is essentially a switch-like device, not just a repeater. It keeps track of the machines on the LocalTalk side of the device, and only relays LToUDP packets for machines it actually has. So this means that if you have two disjoint sets of machines with their own AirTalks, the
aggregate throughput from set A to set B can be substantially
greater than 230 kbit/sec, because the speed of the "backbone" is limited by the speed with which multicast packets can be transmitted over WiFi, not the speed of the line between the Mac and the AirTalk. And latency issues average out the more machines you have on a segment.
But it goes deeper still! This speed, of the WiFi segment, depends on a bewildering number of things, including how crappy the AP is, and how full the RF environment is. The RF bits of WiFi are deep magic and I am not going to pretend I understand them, but they too add their share of chaos to, especially, latency.
So, the tl;dr here is that when you're talking about the speed of a network, you must always be very careful about what you're actually measuring. Marketing materials pick the biggest number and quote that, but marketing is just the empirical mapping of the edges of falsehood, and marketing numbers don't mean very much.
The tl;dr of the tl;dr is: don't use AirTalk because it's fast. It isn't, even compared to wired LT. What it is is more convenient, and that's why it exists.