Re: On removing some obsolete code from subr and core

[Stefan Monnier <monnier@iro.umontreal.ca>]

> you or someone else know better why the default-FOO variables might be
> needed again,

The only reason would be backward compatibility.

> or why they were implemented in the first place.

I'm pretty sure it's for historical reasons.  I don't know those
reasons, tho.  My guess goes as follows:

- First there were normal vars, and buffer-local vars
  (e.g. buffer-file-name, major-mode).
- The two sets were disjoint.
- There was no `default-value' nor `set-default'.
- For those buffer-local vars where a default value was needed/made
  sense, a separate variable was introduced with the convention to name
  it `default-FOO'.

Later on appeared `default-value`, `make-local-variable`,
`make-variable-buffer-local`, ... which blurred the line between those
two disjoint sets of variables, at which point the `default-FOO' vars
became redundant but they were kept for backward compatibility reasons
(also it's occasionally handy to let-bind `default-FOO', while doing the
same for (default-value 'FOO) requires something like `cl-letf').


        Stefan