There was a recent discussion about which lisp/progmodes packages belong in core. A sentiment was expressed that useful languages with non-negligible user bases should probably go in, and others should be in ELPA.
I want to bring up a related point: it should be possible to retire packages from core, once their relevance drops below a critical threshold [1].
I am the former maintainer of the now mostly defunct IDLWAVE mode (lisp/progmodes/idl*.el). IDLWAVE was put in core about 20 years ago, long before ELPA existed, by my predecessor Carsten Dominik (of org-mode fame). I'd advocate moving IDLWAVE to ELPA, for the following reasons:
- The proprietary IDL is a language in rapid decline. In my field (astrophysics), its use is now relegated to older professionals who have not switched to Python (but most of whom wish they had the time to do so).
- IDL's holding company has changed approximately 5 times in the last 10 years, such that the link to its website on Wikipedia isn't even correct.
- None of the free IDL alternatives have really caught on. IDL's costs and licensing restrictions have gotten more onerous over the years.
- IDL is also the "interface description language", confusing users of that unrelated system.
- IDL's file extension ".pro" is quite common, and in use for several other file types, including Qt's UI build system as well as Yarn2/Prolog (which org babel plugin ob-prolog supports).
- Some of the best features in IDLWAVE (e.g. direct documentation linking) required maintenance support from IDL's owners, support which hasn't fully existed for almost a decade.
- The version in core has diverged from the latest (from 7 years ago).
Judging by my email traffic on the topic, there are very few users of IDLWAVE remaining. Even I touch it only a few times a year. No one has stepped up to maintain it in the 7 years since I stepped aside. IDL is still a powerful language and IDLWAVE itself has a great number of features I sorely miss in other emacs programming modes. It deserves support in Emacs. Just not, IMO, in core.
[1] Such a threshold may be hard to define, but here is one idea: if the number of users who activate a mode
on accident is greater than the number who actively seek to do so, it may be time to retire a mode. My experience indicates IDLWAVE has likely passed this threshold.