Grant: NTPsec

In January 2020, ARDC made a $100,000 grant to the NTPsec project at Software
for the Public Interest, which maintains an implementation of the Network Time Protocol (NTP), the Internet protocol for communicating high precision and high accuracy time.

ARDC is not limited to funding amateur-radio related projects; it was created to invigorate both the Internet with ideas from amateur radio, and amateur radio with ideas from the Internet. Yet this NTPsec project is also relevant to hams.

High precision time is used in some ham protocols such as FT8 (Franke and Taylor 8-FSK weak signal modulation). Many hams probably don’t know that FT8 depends on their Internet connection. Digital Mobile Radio also uses accurate time derived from GPS.

With a donation of $100,000, in addition to continuing to pay the lead architect and their apprentice, the NTPsec project will seek out more apprentices to intern, and will focus their attention on developing and testing running NTPsec clock servers over long distance wireless technologies and protocols, with an emphasis on the amateur radio service.

Beyond the amateur radio world, there are bigger issues. GPS-disciplined oven controlled crystal clocks now cost about $90, so many projects are relying on GPS to know accurate time. Yet GPS and related satellite navigation (satnav) has become a single point of failure for a wide variety of worldwide applications. Satnav relies on fragile satellite infrastructure that would likely be attacked, disrupted or jammed during a war. Payment networks and bank clearing systems are rapidly engineering themselves into a state of dependence on a common timebase accurate to a subsecond level. Yet many of the services that would be disrupted by bad time, only work when there’s IP connectivity. So if the service (e.g. payment systems) are up, then NTP service is likely to be up too, even if satnav networks are down.

Some parts of amateur radio share this weak point. If the ham radio “emergency” network stopped working for obscure reasons when satnav stopped working, or when the local connection to the global Internet was down, it wouldn’t be much of an emergency network. And how do you test for that?

ARDC wants to make it likely that reliable and widely deployed NTP implementations can provide online nodes with an excellent backup for when the satnav system gets corrupted or goes down. A solid NTP suite is a necessary condition for that, but not a sufficient one. NTP over Internet is how UTC from national time authorities was propagated for more than a decade, when computer compatible GPS receivers were expensive and uncommon. It’s still readily available and it still works. For example, the US National Institute for Standards and Technology runs NTP servers.

A bit of interesting research would be to measure all the NTP peers on the AMPRnet (and eventually the whole Internet), see where they are configured to get higher stratum clocks from, see what fraction would go down if satellite navigation stopped working, and encourage them to reconfigure for better resilience.

Ultimately the community should consider revising the standard OS distros’ NTP configs so that they specifically include non-satnav clocks that derive from time servers that are solely sourcing their time from earthbound atomic clocks. E.g. make sure that the standard Debian time servers are not relying on satnav, instead relying on Internet connectivity to national standards orgs.

NTPsec hopes to rely on the depth of knowledge that the net44 community already has regarding the latency, jitter, and bandwidth available in existing digital amateur radio networks. Some of the people on the NTPsec team already have various radio operator licenses, and much of the rest can pass the exams within the next test cycle or two.

People from ARDC are invited to join the NTPsec developers’ mailing list and describe their
stories and needed features, and we all can argue and reason together, think hard together, and work out the work ahead.