Docker image deprecation
For each new LarvaTagger version, two additional Docker images are pushed to Docker Hub. Docker images are made persistently available to ensure reproducible tagging results and - to a lesser extent - user experience.
Indeed, although the build recipe is also available, building an image using the same recipe will likely result in a different image. Many packages are not version-locked, including OS packages installed using apt
and most Python packages installed using poetry
. In addition, images have been built using git heads for LarvaTagger.jl, TaggingBackends and MaggotUBA-adapter. The corresponding commits or tags can be easily guessed, especially as they have been documented in the change logs, but one has to adapt the recipe.
These Docker images accumulate and amount to several tens of Gb on Docker Hub. While this may still be an acceptable repository size, several images have never been pulled, not even once. Since the latest version is to be favored for any new project, I am considering archiving and removing these unused images from Docker Hub.
The tags considered for removal/deprecation currently include:
- 0.13.1-standalone
- 0.13-standalone
0.12-20230129- 0.12-standalone
- 0.11.1-standalone
- 0.11-20230129
- 0.11-standalone
- 0.10-standalone
- 0.9-20230129
- 0.9-standalone
- 0.8-standalone
- 0.7.2-20221228
0.7.2-20221005- 0.7.2-standalone
- 0.7.1-20221005
- 0.7.1-standalone
- 0.7-standalone
- 0.6.1-20221005
- 0.6-standalone
- 0.5-standalone
Note(s):
- Unused versions 0.7.2-20221005 and 0.12-20230129 will not be removed/deprecated in the first wave, because they are the last images that shipped with tagger 20221005 and 20230129 respectively.
As a consequence of being removed from Docker Hub, archived versions would have to be expressly requested to me (@flaurent).
A second wave of deprecations may also be planned at some point, with at least 6-month notice, to remove the oldest images that are suspected to have never been used to generate data or published results. The oldest known pipeline is t5_analysis_replicates (dormant) and is based on version 0.11.1.
The first wave of deprecations is not planned yet and should not happen before the 1.0 release anyways. This is a "pre-announcement" to give time for comments or requests, if any.