I thought about trying this a while back, however all the solutions
looked *very* ugly.
Assuming that the next major release of CT will be Windows only, it would
be very useful if CT was made WINSOCK aware. What I envision is having
each machine attach to a multicast address/port to which new log entries
and cluster spots would be sent. This could be used instead of the current
serial network. (I'm not advocating dropping support for the serial network.)
There are some advantages to this:
1) RG58 Ethernet has better immunity to high RF environments
than typical RS232 cables. (This from experience at VE7ZZZ
with 6 bands @ 2KW each.)
2) Configuring a PC with more than two serial ports is difficult
at best. To fully automate a station under windows requires
five serial ports: mouse, radio, network in, network out,
and rotator. You can drop this to four if you use a bus mouse.
3) A central server could attach to each workstation via a TCP
connection and collect all logging data in one place. TCP ensures
nothing is lost, and it's immediately apparent when the server
or client goes away, making recovery simple to implement.
4) If a workstation in the middle of the network goes away you
don't partition your network.
It's interesting this was brought up now - this week I'm going to be
wiring the VE7ZZZ site for Ethernet (RG58) and serial (for CT). Let me
explain what my goals are.
The VE7ZZZ group consists of a core group of five or six die hard
contesters, and 20 or so locals who come out to operate casually for
a few hours on Saturday or Sunday. The core operators generally provide
the computers, and each has his own copy of CT. The computers are all
standalone, and range from Heathkit 8088's to i486 DX/2-66's. I don't
think any two machines have the same version of CT. Every machine also
seems to have it's own unique combination of size and density for floppy
drives. This results in a total lack of consistency between operating
positions, and merging logs after the contest is an adventure in
frustration. The lack of networking for CT really hurts us during
the WPX contest since we cannot coordinate the millions of multipliers
for all the US prefixes :-)
There are four areas that have to be addressed:
1) CT must be networked so that we can maximize our scoring
2) Consistent interfaces must be presented to the operators
3) We cannot rely on operator supplied computers to have
the necessary resources for proper operation (i.e. RAM,
disk space, serial ports, software)
4) Since many operators arrive very close to, or after, the
start of the contest (00Z is 1600 local, 1700 in the summer)
it must be possible to get these computers operational as
quickly as possible
The solution we are implementing is a two stage process. First, we are
installing serial cables to allow us to use CT's network features. This
will benefit our scoring immediately, and allow the operators to become
familiar with the advanced CT features we aren't currently using.
The second stage involves installation of Ethernet throughout the site.
A server running FreeBSD will also be installed. All software used at
the site will be centralized on the server and exported to the workstations
using NETBIOS. Ethernet cards will be purchased for each operating
position. These cards will be equipped with BOOTP ROMs. When a machine
is brought in the setup will consist of installing the Ethernet card,
attaching it to the RG58 drop, and booting from the network. This will
boot the machine directly into Windows for Workgroups, and the appropriate
filesystems will be mounted from the server using NETBIOS over TCP.
The Windows environment will provide the following software in a
standard configuration:
* CT (latest stable release)
* WF1B RTTY software (latest stable release)
* Grayline software
* Propagation analysis and prediction
* Access to latest solar reports (from STD, hopefully with a
graphical display)
* Interactive messaging between workstations
* Satellite tracking
While this solves our problems with setup time and consistency, we still
fall short on serial ports - three are needed to run windows and attach
to the network. One option would be to write a TSR that hooks INT14 and
redirects COMx I/O over a TCP connection to the BSD server, allowing us
to fake the CT serial network. The BSD end is simple to implement, however
I'm not familiar enough with DOS, WIndows, or CT to consider this a trivial
task (or even know if it's possible, although Beame&Whiteside does provide
an INT14 hook TSR with their DOS networking product that lets you use things
like Procomm to establish telnet connections).
It would make the most sense to have CT use the network directly to
establish connectivity between the workstations. CT currently assumes
an unreliable network. This makes it very easy to map the semantics
of the current network to IP multicast. It would be necessary to assign
a multicast IP address for CT's use. It would also be necessary to
define some "well known" ports for CT's use:
BCAST All stations broadcast
CLUSTER Packet cluster spots
BASE Station 1
BASE+1 Station 2
.
.
.
BASE+n Station n
By using ports to address the stations it isn't necessary to configure
individual workstations with the IP addresses of all the other workstations.
It also makes it possible to write software to monitor the operation of
the entire network, allowing band captains and station managers to analyze
trends in real-time, and make suggestions to the operators (band changes,
mult vs. run, etc) to increase their point rates.
Assuming the above scenario, the PC at each operating position looks like
this:
* i386 or better CPU
* 8 MB of RAM
* local hard disk with 2 MB free space
* one serial port for mouse
* (optional) serial port for radio connection
* one free ISA slot for Ethernet card
* (optional) parallel port for CW keyer or RTTY
This eliminates one of our PC's, however that's not a bug :-)
The costs to implement this are surprisingly low. We can piece the server
together for well under $1000 (i486 DX/2-66, 12 MB RAM, about 1 GB of disk,
3C509 Ethernet, dot matrix printer). Ethernet NE2000 clone cards can be
had locally for about $60, and BOOTP ROMs are about the same price, so we're
looking at 6X120=$720 to equip the operating positions. The station
infrastructure costs are:
Conduit, wall boxes, elbows $150
BNC's, wallbox plates 50
BNC's & tee's for workstation drops 50
I haven't included the cost for CT and WF1B site licenses as these
are currently unknown. Regardless, we should be able to do the whole
thing for about $2000 (Canadian).
My apologies for rambling all over the place. I think the context was
necessary to show the not-so-obvious benefits of making CT IP network
aware.
--lyndon VE7TCP
--
Submissions: ct-user@eng.pko.dec.com
Administrative requests: ct-user-request@eng.pko.dec.com
Problems: reisert@eng.pko.dec.com
|