pgRouting vs Apache OpenNLP

The only visible activity is the pgRouting 4.0.0 pre-release chain: alpha1 in September, beta1 in October, rc1 in November. Alpha1 carries all the substance — the C++ standard moves to C++17, documentation for all four languages builds by default, and the function catalog gets a systematic pass in which output columns are standardised to a common shape and combinations signatures are promoted from proposed to official across the shortest-path family. The beta and candidate that follow only correct documentation about which functions are experimental, proposed or official.

This is a major version being spent on consistency rather than new algorithms. Standardising outputs to (seq, path_seq, start_vid, end_vid, node, edge, cost, agg_cost) across functions that previously each had their own column set is the kind of change that only happens at a major boundary, and promoting combinations signatures to official settles an API that has been provisional for years. The two releases after alpha1 changing nothing but function-status documentation suggests the classification of experimental versus official is itself the contested part of this release.

Expect 4.0.0 final to follow with little functional change beyond what alpha1 established, since both subsequent pre-releases were documentation-only. The absence of any release after rc1 in November is the notable gap — nothing in these entries explains it.