Subject: [linux-audio-dev] Re: ethernet vs serial MIDI again ...
From: jfm3 (jfm3_AT_acm.org)
Date: pe tammi 21 2000 - 10:26:11 EST
Benno Senoner wrote:
> On Thu, 20 Jan 2000, jfm3 wrote:
> > Dan Hollis wrote:
> >
> > > On Thu, 20 Jan 2000, jfm3 wrote:
> > > > 10/100baseT doesn't make a good real time control protocol physical layer. It's
> > > > not "real time" enough.
> > >
> > > And 38400 baud aysnc serial is? Boggle.
> >
> > Don't think about throughput, think about latency. When I send a message out on a
> > MIDI cable, I have 100% certainty of when it will arive at its destination(s) within
> > constant bounded tolerances (1 ms.). Similar claims can not be made of ethernet.
>
> I am not an ethernet expert, but assume that you run 10Mbit (or 100Mbit) through
> a switch, which eliminates the collisions.
>
> Of course if there is a trasmission error or packet loss, this packet will be
> delayed and retransmitted. [...] any thoughts ?
I suppose if one worked at it long enough one could hack an ethernet connection to get
real time control signal transmission with acceptable error rates. I guess my point is
not about what is possible, but that in what I've read of efforts to build a better MIDI,
the consensus has been to use a cleaner design for the physical transport than ethernet.
I'd rather not have to explain the real time performance differences between a "switch",
a "hub", and an "ethernet bridge" to the sales enemy at Sam Ash, for example.
(jfm3)
This archive was generated by hypermail 2b28 : pe maalis 10 2000 - 07:23:26 EST