Skip to content

Testing and interop

The open IEC 103 ecosystem is lopsided: there are several controlling systems, and only one open relay that serves test data, which deviates from the standard. So the protection equipment side of go-iec60870-5 has strong independent evidence and the controlling side has little.

go-iec60870-5 tests every ASDU against the examples of the standard’s annex B, round trips and fuzzing, runs the client against the server over in-memory and lossy lines, and adds hostile-peer and chaos tests. Recorded traffic of an SEL RTAC polling several relays decodes and encodes back unchanged.

iec103ctl runs every command against go-iec60870-5’s own relay and controlling system (make e2e in iec60870ctl), and against the reference images below.

iec60870-5-interop v0.5.0 packages four independent IEC 103 implementations, unbalanced, over the FT 1.2 byte stream on one TCP connection:

Adapter Implementation Language Roles
vitusdong vitusdong/iec103-dotnet v0.1.0 C# server, client
ariec103 masarray/ARIEC103 C# client
open103 XVincentX/Open103 C++ client
goiecp5 riclolsen/go-iecp5 v0.4.7, package cs103 Go client

Images are ghcr.io/otfabric/iec60870-5-interop-iec103-<adapter>. Every reference server deviates from the standard in ways listed in the suite’s known-bugs.tsv: vitusdong answers an interrogation with cause 11 and a shortened ASDU 1, confirms commands with cause 11 and identifies itself without a compatibility level. go-iecp5 cs103 is the most standard-conforming controlling system of the survey.

ghcr.io/otfabric/iec60870-5-interop-iec103-otfabric is go-iec60870-5 behind the same container contract as the reference images (adapter name otfabric), new with iec60870-5-interop v0.5.0, so you can test your IEC 103 controlling system or protection equipment against it, the way OT Fabric tests its own stack against yours:

Terminal window
docker run -d --name go-iec103 -p 127.0.0.1:2401:2401 \
ghcr.io/otfabric/iec60870-5-interop-iec103-otfabric:<tag> server
until docker exec go-iec103 test -f /run/iec60870-5-interop/ready; do sleep 0.1; done

The relay serves the baseline fixture on port 2401, with generic services and disturbance data, and client <operation> runs go-iec60870-5’s controlling system against your relay. Against a relay other than vitusdong, the client bugs of vitusdong and Open103 listed in the suite’s known-bugs-consumer.tsv show up on the wire; apply that list with known-bugs.tsv when you test against it.

Its tags are go-iec60870-5 versions, such as v0.5.1, and latest; a rebuild of a version replaces its tag, so pin the digest when you need an exact build, and version prints the versions inside the image. It is not part of an iec60870-5-interop release or its manifest.json, and it reports independentEngine: false: a result against it tells you about your stack, not something independent about OT Fabric’s. A result you publish, naming your implementation and version, is independent evidence. Like the references, it is for testing only and has no place on a network you do not control.

make interop-iec103 runs against the pinned images of iec60870-5-interop v0.4.0. The last recorded run (2026-11-12): 75 subtests passed, 38 skipped as known bugs or undeclared features, none failed.

Direction Result
Reference clients against the go-iec60870-5 relay goiecp5 15 of 15, Open103 15 of 15, vitusdong 14 (one feature it does not declare), ARIEC103 9 (six known client differences: commands without RII, clock synchronization with CP32Time2a)
Disturbance data transfer against the go-iec60870-5 relay Open103’s control system runs the whole procedure; its value decoding is a known upstream bug
go-iec60870-5 client against vitusdong Link start-up, clock synchronization, GI termination and an idle line pass; identification, GI answers and command confirmations are known server bugs

iec103ctl v0.9.1 passes 73 checks against the same four stacks; the skips are known upstream bugs and features an image does not declare.

  • A standard-conforming relay. No open stack offers one, so the client has one deviating peer. Reports from real relays, with the --debug trace, are welcome.
  • Generic services against an independent peer. vitusdong is the only stack with them, and it uses INF 0 in every generic ASDU; ARIEC103, Open103 and go-iecp5 cs103 have none.
  • Independence. vitusdong’s tests use a copy of the ARIEC103 slave simulator as their oracle, so those two are not fully independent.
  • Shared lines with several relays polled by one client: not implemented.