Philip McGrath

Substantively, this seems like a good change to me. Changing the default for `current-alist-separator-mode` would, strictly speaking, be a backwards incompatible change. That may be the right choice for all...

I was trying to split the minimal changes into a different commit than the broader reformatting, which involved rewriting the history. (Maybe there is a more elegant way to do...

Actually, as I read `editor:backup-autosave-mixin` more closely, it looks like it always calls `path-utils:generate-autosave-name` or `path-utils:generate-backup-name` immediately before attempting to save the file, so I guess the thing to do...

Reading more from the Reference on [Windows paths](http://docs.racket-lang.org/reference/windowspaths.html), it looks like Racket already handles avoiding `MAX_PATH` by switching to the `\\?\` form: > Outside of Racket, except for `\\?\` paths,...

I like the idea of eliding some bytes from the middle: in most cases the resulting name would probably still be decipherable to the user (certainly more so than a...

I've added tests to confirm that `path-utils:generate-autosave-name` and `path-utils:generate-backup-name` fall back as expected when the directory from the relevant preference is deleted. I was unsure whether I should also add...

Do you think we should check that the current user has write permission in the same places as `directory-exists?`? I can see an argument for not checking and letting the...

About checking write permission: this is my ignorance about the preference system, but, if the predicate function given to `preferences:set-default` reports that a value from the user's preferences file is...

Ok, I have dealt with permissions, added tests for both that and long file names, and updated the docs accordingly. I've also edited the history back down to two commits.

While running tests, I found two more todos: 1. Currently, my tests appear to overwrite the user's values for these preferences. I had thought I had avoided that by using...