rirdb

Daily merged Regional Internet Registry (RIR) allocation data with reproducible release artifacts and a semantic diff between consecutive builds.

The repository combines the extended delegated statistics published by:

• APNIC
• ARIN
• LACNIC
• AFRINIC
• RIPE NCC

A GitHub Actions workflow builds a new snapshot every day and publishes it as a GitHub Release.

Daily build process

The workflow is defined in .github/workflows/daily-release.yml and runs every day at:

03:17 UTC

It can also be started manually with workflow_dispatch.

Each run performs the following steps.

1. Determine the release date

The workflow uses the current UTC date and creates a release tag in this form:

rir-YYYY-MM-DD

Example:

rir-2026-10-21

2. Capture the previous published build

Before generating the new snapshot, the workflow finds the dataset that should be used as the comparison baseline.

For a normal daily run, it selects the most recently published earlier rir-* release.

For a rerun on the same UTC day, if that day's release already exists, the workflow downloads the currently published dataset from that same release before rebuilding it. This means a same-day rerun produces a diff against the build it is about to replace.

The previous release asset used as input is:

delegated-rir-extended.txt.gz

It is decompressed before comparison.

If there is no previous release, the build still succeeds and produces an empty diff marked with:

previous=none

3. Download all five RIR datasets

scripts/merge-rir.sh downloads the current extended delegated statistics from:

Registry	Source
APNIC	https://ftp.apnic.net/apnic/stats/apnic/delegated-apnic-extended-latest
ARIN	https://ftp.arin.net/pub/stats/arin/delegated-arin-extended-latest
LACNIC	https://ftp.lacnic.net/pub/stats/lacnic/delegated-lacnic-extended-latest
AFRINIC	https://ftp.afrinic.net/pub/stats/afrinic/delegated-afrinic-extended-latest
RIPE NCC	https://ftp.ripe.net/pub/stats/ripencc/delegated-ripencc-extended-latest

Downloads use retries, connection timeouts, and a maximum transfer time.

4. Validate each RIR input

Each downloaded file is checked before it can enter the merged dataset.

The validation requires records whose registry field matches the expected RIR and whose resource type is one of:

ipv4
ipv6
asn

The build fails if a source does not contain a minimum expected number of valid delegated records. This prevents an empty response, error page, or obviously incomplete upstream file from being published as a valid snapshot.

5. Merge the delegated records

Only IPv4, IPv6, and ASN records are retained.

The merged file begins with:

# registry|cc|type|start|value|date|status|opaque-id

Records then retain the delegated-statistics field layout, for example:

ripencc|PL|ipv4|91.198.90.0|256|20100115|allocated|...
apnic|JP|ipv6|2001:200::|32|20020701|allocated|...
arin|US|asn|14618|1|20051104|allocated|...

The primary merged artifact is:

delegated-rir-extended.txt

A deterministic gzip representation is also produced:

delegated-rir-extended.txt.gz

The gzip output is created with metadata disabled so identical input produces identical compressed output.

6. Count resources in the snapshot

The build writes:

counts.txt

with totals for each resource type:

ipv4=<count>
ipv6=<count>
asn=<count>

7. Generate the consecutive-build diff

scripts/diff-rir.sh compares the previous merged snapshot with the current one.

The comparison is set-based rather than line-position-based:

1. comment lines are removed;
2. records are sorted;
3. duplicate records are collapsed;
4. the two normalized sets are compared.

This prevents upstream record ordering from appearing as a data change.

The resulting file is:

changes.txt

with a header such as:

# previous=rir-2026-10-20
# current=rir-2026-10-21
# format=operation|registry|cc|type|start|value|date|status|opaque-id

Each changed record starts with an operation:

-|ripencc|PL|ipv4|...
+|ripencc|PL|ipv4|...

where:

• - means the complete delegated record existed in the previous build but not in the current build;
• + means the complete delegated record exists in the current build but not in the previous build.

A modification to any field is represented as one removed record and one added record.

A compressed form is also published:

changes.txt.gz

8. Count changes

The build generates:

changes-counts.txt

with aggregate change counts:

added=<count>
removed=<count>
added_ipv4=<count>
removed_ipv4=<count>
added_ipv6=<count>
removed_ipv6=<count>
added_asn=<count>
removed_asn=<count>

These counts are also included in the GitHub Release notes.

9. Generate checksums

SHA256SUMS is regenerated after all artifacts have been created.

It covers:

delegated-rir-extended.txt
delegated-rir-extended.txt.gz
counts.txt
changes.txt
changes.txt.gz
changes-counts.txt

Consumers can therefore verify the integrity of every published data artifact.

10. Publish or update the GitHub Release

If the daily release does not yet exist, the workflow creates it.

If the release already exists, such as during a manual same-day rerun, the release notes are updated and all assets are uploaded with replacement enabled.

Each release therefore represents the latest successful build for that UTC date.

Release artifacts

Every daily release contains:

Artifact	Description
delegated-rir-extended.txt	Full merged APNIC, ARIN, LACNIC, AFRINIC, and RIPE NCC snapshot
delegated-rir-extended.txt.gz	Deterministic gzip-compressed snapshot
counts.txt	IPv4, IPv6, and ASN record counts
changes.txt	Semantic record diff against the preceding published build
changes.txt.gz	Compressed diff
changes-counts.txt	Added and removed record counts by resource type
SHA256SUMS	SHA-256 checksums for all published data artifacts

Running locally

Build a fresh merged snapshot:

bash scripts/merge-rir.sh dist

This produces the base snapshot artifacts under dist/.

To compare two locally available merged snapshots:

bash scripts/diff-rir.sh previous.txt current.txt dist previous-label current-label

For example:

bash scripts/diff-rir.sh \
  snapshots/rir-2026-10-20.txt \
  dist/delegated-rir-extended.txt \
  dist \
  rir-2026-10-20 \
  rir-2026-10-21

Diff semantics

The diff describes changes to complete delegated-statistics records, not inferred business events.

For example, if an IPv4 allocation changes status, country code, date, size, or another field, the old full record is emitted with - and the new full record with +.

The diff therefore answers:

Which delegated records disappeared, and which delegated records appeared, between two consecutive published builds?

It does not attempt to infer why a registry changed a record.

Data interpretation

The merged database reflects delegated statistics published by the five RIRs. It should be treated as registry/allocation data rather than as an IP geolocation database.

In particular, the country-code field is part of the delegated registry data and should not be assumed to represent the current physical location of every address in the corresponding resource.

Failure behavior

The workflow is intentionally fail-closed for source acquisition and validation.

A daily release is not published if:

• a required RIR source cannot be downloaded;
• a download is empty or obviously incomplete;
• a source fails the delegated-record validation;
• merge or diff generation fails;
• GitHub Release publication fails.

The previously published release remains available if a later run fails.