Liblognorm is a fast-samples based normalization library. It’s brand new version 1.0.0 will be released today. It is a major improvement over previous versions, but unfortunately we needed to change the API. So some notes are due.
Liblognorm evolves since several years and was intially meant to be used primarily with the Mitre CEE effort. Consequently, the initial version of liblognorm (0.x) uses the libee CEE support library in its API.
As time evolved, the initial CEE schema underwent considerable change. Even worse, Mitre lost funding for CEE. While the CEE ideas survived as part of Red Hat-driven “Project Lumberjack”, the data structures became greatly simplified and JSON based. That effectively made libee obsolete (and also in parts libestr, which was specifically written to support CEE’s initial requirement of embedded NUL chars in strings).
Recently, Pavel Levshin converted liblognorm to native JSON, which helped improve performance and simplicity for many client applications. Unfortunately, this change broke interface compatibility (and there was no way to avoid that, obviously…).
The current library is the result of that effort. Application developers are encouraged to switch to this version, as it provides the benefit of a simpler API. This version is now being tracked by the git master branch.
However, if you need to stick to the old API, there is a git branch liblognorm0, which contains the previous version of the library. This branch is also maintained for important bug fixes, so it is safe to use.
We recommend that packagers create packages both for liblognorm0 and liblognorm1. Note that liblognorm’s development packages cannot coexist on the same system as the PKGCONFIG system would get into trouble. Adiscon’s own packages follow this schema.
Note that rsyslog will soon begin to enjoy the benefits of liblognorm1. This results in a notable performance improvement for mmnormalize. Support will initially become available in v8.