On 3/12/2010 11:08 AM, John Hays wrote:
I would avoid the complexity of domain names as you outline here.
The bandwidth of the D-STAR DD stream doesn't bode well for supporting
"server" applications at the "last mile" ID-1.
If you mean putting the server application at the ID-1 end of things,
absolutely right. That'd suck.
I was speaking of putting that functionality in the
Gateway/Router/whatever you want to call it.
(You're saying you're "creating" the concept of a DD Router, when that's
exactly what the Gateway is doing with DD packets already -- just a very
poor implementation. I would NOT want an ADDITIONAL fragile PC up at
the repeater site to handle DV and DD separately. That's a poor use of
both hardware resources (two machines where one can easily do this job)
and human resources (two machines will break twice as often, and require
site visits commensurately - doesn't follow the KISS principal.)...
To be deadly honest, DD bandwidth is almost too low in today's world
already to be of much usefulness for *anything* modern IP networks are
used for. It's barely even fast enough for "okay to use in a pinch". I
had 128Kb/s FULL-duplex over ISDN delivered to my house in the early
90's, cheap.
It's 20 years later and 128 Kb/s half-duplex with relatively high
turn-around time on the TX/RX switching, is far too slow for something
that has an Ethernet device plugged into it.
It's a $1000 low-speed data play-toy, from the viewpoint of anyone who's
done wireline data work. We buy fiber like there never was copper in
the ground for most applications these days...
Just my not-so-humble opinion... keep DD really really dirt-simple,
because you put five ID-1s on the air in an area simultaneously and all
pull up a web page at the same time, the air interface is well beyond
saturation, through-put-wise. I'd love to see the results of a
through-put test with one ID-1... two ID-1s... three ID-1s... four
ID-1s... all using the same data module and gateway. Also try it with
all five able to "see" each other via RF paths directly, then place one
behind a ridge-line and watch the network barf from "hidden node
syndrome", something that was very well documented in the early 90's in
1200 and 9600 baud Packet Radio...
This is why Packet ended up having it's most "popular" and "useful" mode
really being APRS... a "fire-and-forget" UDP-like system that doesn't
care if the other nodes in the network actually received the data.
Collisions kill multi-node RF data systems unless there's a central
time-slot "traffic cop" at the highest/most visible location in the network.
Some of the old Packet folks could easily describe (or maybe even find
notes on) how to do these throughput measurements. I suspect, as I said
above, that five simultaneous ID-1's trying to pull and push data, would
completely crush the local on-air "LAN", and render it useless and/or so
slow, as to be so.
A five node LAN ain't all that great shakes. Even if it performed well
up through ten nodes (highly doubtful)... that's still a very wimpy data
LAN these days.
Nate WY0X
