Juhan Leemet wrote on Wed, 01-Oct-2003: > >> As cool as that sounds, there is something really weird about having > >> something with as much latency as the internet connected to something > >> as low-latency and sample accurate as JACK. > > > >Just think of it as a delay line :) > > I think the biggest problem is that it is a (randomly?) _variable_ delay line. > Some early voice-over-internet work was done more than 20 years ago, and it > was deemed "not ready for prime time" mostly because of lack of "guarantee of > service/quality" (i.e. variable latency, and occasional dropout). The > underlying technology has not changed (that) much (if at all?). TCP/IP was > defined in RFCs when? 20 or 30 years ago? Some refinements, granted. Yes, but given some parameters of your intended network you *should* be able to pick a fixed latency that will be big enough almost all of the time. Obviously, this isn't foolproof, and it probably doesn't scale, but given a controlled network it might just work well enough for some applications. When it doesn't, well... try something else. For instance, routing audio from a recording/mix machine in your studio to another in a different room over a 100/1000 Mb ethernet link for the purposes of mastering. The network is controlled, fast, *and* the latency doesn't particularly matter in this context. Now try to do this over the internet generically, and you'll have to settle for both massively increased latency and reduced channel count, and you'd probably need to venture into lossy compression too (as an option) if that satisfies your needs. Anyway, it will be a fun experiment. jlc
