Projects Spectrum Next NetBSD Retrocomputing

A time server for the ZX Spectrum Next

Published · 3min

I recently received my Spectrum Next. I’d been waiting three years for it, and they started sending them out just as the lockdown started. Thankfully, even though I wasn’t able to update the address, they were able to receive it at the office, and drop it over, which relieved me no end.

The model I got was the one with the RPi Zero Accelerator board, WiFi, and the real-time clock. It’s been fun to play with so far. I had a C64 as a kid, but I had a bit of a soft spot for the Spectrum, even if it wasn’t quite as capable as the C64 for the most part, and the Spectrum Next looked like a really fun project.

This isn’t about any of that though. As mentioned, my model has the real-time clock add-on. Natively, NextZXOS supports a protocol called the Network neXt Time Protocol. Something like the Spectrum Next doesn’t require anything better than second-level accuracy, and something like NTP, or even SNTP are much too complicated for a device as small as the Next, so NXTP keeps it as simple as possible: you the server for the current date and time in a given timezone and it responds with date and time. No integrity beyond a very simple checksum, totally cleartext, but it’s good enough.

Unfortunately, the default server implementation is written with .NET Core, and I wanted to be able to install this on a Raspberry Pi sitting in my flat running NetBSD 9, so I knocked together my own server and client implementation in Go. When built, it weighs in at just over 2MB. Were I to write it in C, I could probably make it significantly smaller, but there’s not much point in that, and it’d just complicate the build for anybody who wants to use it.

Hopefully, it should be of use of other Spectrum Next owners.

Running it under NetBSD

What’s left is to run it as a daemon on the Pi, which is easier said than done with servers written in Go, as this is apparently very finicky. The alternative is to have something else daemonise it. On FreeBSD, the daemon command would do the trick, but NetBSD lacks it. However, daemond fits the bill:

$ sudo pkgin install daemond

Unlike FreeBSD’s daemon command, daemond provides no way to save the PID of the child process it spawns, so we’ll have to use pgrep to figure it out, which is less than ideal. Here’s the rc script I wrote:

# PROVIDE: nxtp

if [ -f /etc/rc.subr ]; then
    . /etc/rc.subr

nxtp_flags="-endpoint :12300"

nxtp_start () {
    echo "Starting ${name}."
    /usr/pkg/bin/daemond ${command} ${nxtp_flags}
    /usr/bin/pgrep ${name} > ${pidfile}

if [ -f /etc/rc.subr -a -f /etc/rc.conf -a -f /etc/rc.d/DAEMON ]; then
    load_rc_config $name
    run_rc_command "$1"
    case ${1:-start} in
        if [ -x ${command} ]; then
        if [ -f ${pidfile} ]; then
            pid=$(/usr/bin/head -1 ${pidfile})
            echo "Stopping ${name}."
            kill -TERM ${pid}
            echo "${name} not running?"
        $0 stop
        sleep 1
        $0 start
        if [ -f ${pidfile} ]; then
            pid=$(/usr/bin/head -1 ${pidfile})
            echo "${name} is running as pid ${pid}."
            echo "${name} is not running."

This went into /etc/rd.d, and nxtp went into /usr/local/bin. The final bit is to add nxtp=YES to /etc/rc.conf and run sudo service nxtp start.