One of the projects on the project page is to give all Octave errors/warnings an ID.  Would it be sensible to try to harmoinze Octave and Matlab IDs at the same time?

I'm not sure if TMW has copyrighted their message IDs, but for short ones (like singular matrix) that are presumably too short to copyright, it would nice for Octave's ID to be the same as Matlab's ID, subject to the "Octave:" prefix and spinal-case replacing camelCase.

For the project on adding IDs, would the goal be to have a small set of IDs used in multiple places, or for each customized error to have its own ID?  There are already different error messages that have the same ID (like for non-conformable operands, where the message indicates the operator and the operand sizes).  How far should that extend?

just for the record, Matlab's are CamelCase, e.g., 'MATLAB:UndefinedFunction' vs. 'Octave:undefined-function' (real examples).

I think users could accommodate different ids, e.g.:

    octave:37> ones(2) \ [1;1];  [msg, id] = lastwarn ()
    msg = warning: matrix singular to machine precision, rcond = 0
    id = Octave:singular-matrix-div

provided they were always non-empty, unlike this:

    octave:36> inv(ones(2));  [msg, id] = lastwarn ()
    msg = warning: inverse: matrix singular to machine precision, rcond = 0
    id =

it'd be pretty widespread change though...