I guess the updated list of functions is a bit orthogonal to the original issue. I think there's some confusion here about the goal of the error messages displayed and I'd argue that this is i npart working as intended. With the statistics package unloaded, we get:

1. for nlinfit a message saying to load the optim package. The nlfint function is part of the Matlab optimisation package, so Octave says that it is probably in the optim package. It does not know whether it is implemented or not, but if it, then it should be done there.

2. for linprog, it tells to use glpk. Besides pointing to the right package (for Matlab functions), it also has some smarter hints for alternatives. Maybe it's time we stop recommending glpk for those that want to use linprog.

3. for powell we just have standard missing function error. powell is not a function in any of the Matlab toolboxes, that's why there's no "smarter" error message. I don't think we want to compile all the functions for all the packages (and deal with possible clashes).

It would only take a minute or so to tweak the three optim function cases mentioned in comment #0, but looking at the optim function list on Octave Forge [1] (which I believe hasn't been updated in years) there are 55 functions listed while the __unimplemented__.m file on default lists about 20. So, there definitely seems to be a need to at least bring it up to date, likely the same for all of the packages. Ideally it would be brought up to date with the latest releases. Could bring it up to date with the out-of-date Octave Forge manually with some scripting or clever copy/paste/reformatting from the website. But I also noticed that there are a lot of package function names in __unimplemented__.m that haven't been implemented yet.  It's not clear how the list of functions that 'should be' in the package was generated. Just copy/pasting the list from Octave Forge, we'll lose all of those.

Does anyone know where those came from? Any idea how to update a package function list but distinguish between those 'not yet' functions from those that may have been removed or moved elsewhere?

[1] https://octave.sourceforge.io/list_functions.php?sort=package#optim

Echoing Markus, it should be easy to simply move functions between the packages that they actually exist in within the script _unimplemented_.m

If the optim package is unloaded, I get

help nlinfit
error: help: The 'nlinfit' function belongs to the statistics package from Octave

Interestingly, for another function from the optim package I get

help linprog
error: help: Octave does not currently provide linprog.  Linear programming problems may be solved using 'glpk'.

And this one:

help powell
error: help: 'powell' not found

Quite confusing.