The story
Licence plates, a satellite called SARAMAGO, and why old software goes quiet. No jargon, with a glossary.
The plain-language version of the migration, for colleagues who do not parse TLEs for a living.
The world ran out of five-digit satellite numbers. Does your software know?
On 11 July 2026 the satellite catalog reached number 100000. Much older software can quietly lose every satellite catalogued since. This is a free test kit that shows developers exactly where their code breaks.
A conformance corpus for parsers crossing the five-digit catalog-number boundary
Sixteen cases built from real CelesTrak responses and published specification vectors: Alpha-5 TLEs, six- and nine-digit NORAD_CAT_IDs, and CCSDS OMM in CSV, JSON, XML and KVN. Every expected value traces to a recorded URL, retrieval time and SHA-256. Point the runner at your parser; it tells you, case by case, what breaks and why.
What the catalog looks like todayTracker snapshot
Highest satellite number seen in recent launchesHighest catalog number, last-30-days group
100,789
2026-09-21 05:51 UTC
Recent satellites with six-digit numbersSix-digit objects in that group
231 of 231
2026-09-21 05:51 UTC
Old format asked for recent satellitesTLE request for that group
404
“not found”: none of them fitNo GP data found · 2026-09-21 05:51 UTC
Nine-digit numbers in Starlink launch dataNine-digit ids in the Starlink SupGP feed
yes (27)
2026-09-21 05:51 UTC
Counts and yes/no states only, computed once a day by a scheduled job; this page never contacts CelesTrak. Full tracker with history.
Where to go next
The migration
The full technical version: formats, column layouts, the Alpha-5 letter table and what CelesTrak actually returns.
Alpha-5, OMM, what CelesTrak serves, a TLE-versus-OMM side-by-side on the CCSDS example, and what to change.
Findings
What broke when three different programs read today’s data. One bug we found and reported to a popular free library, with a fix we submitted; a second bug, reported by someone else, whose pending fix by another contributor we tested against the corpus.
Naive versus reference versus python-sgp4 across sixteen cases, the CelesTrak TLE 404, provider rendering differences and the upstream reports.
Tools
Turn a satellite number into its new lettered form and back, and take a TLE apart field by field. Everything runs in your browser.
A strict Alpha-5 encoder/decoder checked against the specification vectors, and an interactive TLE anatomy with a paste box. Client-side only.
Library status
Every week we check whether two popular free libraries handle the new numbers yet.
Weekly no-network reproducers against the latest PyPI python-sgp4 and Skyfield, with upstream links.
Activity
The bug reports and fixes sent to the python-sgp4 library, the test kit’s releases, and a timeline of milestones.
Upstream issue and pull-request state from the GitHub API (edge-cached hourly), the corpus's latest release, and a hand-maintained timeline.
How it was built
Where the numbers come from, who checked them, and who is responsible.
Provenance, the independent audit, AI assistance under human direction, NEOGY LLC and the maintainer.
The corpus
- gp-omm-conformance on GitHub — version 0.1.0, MIT, audited before publication.
- Concept DOI 10.5281/zenodo.22867654 (always the latest release); version DOI 10.5281/zenodo.22867655 for exactly v0.1.0.
- python-sgp4 issue #171 — a bug we found and reported: the library crashes on records that have no international designator.an empty
<OBJECT_ID/>in OMM XML makessgp4.omm.initializeraiseTypeError; reported from the corpus. - python-sgp4 PR #172 — the fix for that bug, submitted at the maintainer’s invitation and awaiting review.the fix for #171 (two files, two tests), submitted at the maintainer’s invitation, awaiting review.