RE: File modes facilities.

["Drew Adams" <drew.adams@oracle.com>]

    > Yes, if you type the directory explicitly and absolutely (no
    > `../', `./', `~/'), then `load-library' will complete in that
    > directory.  My point was that, without doing that, they complete
    > differently - they use different paths (directories).

    Yes, they do work differently.  I haven't advocated to remove load-file
    and/or merge it into load-library for precisely this reason.
    But we could get rid of load-library if we could just add an interactive
    spec to `load'.

    As for getting rid of load-file, maybe you could make load-library's
    completion understand a leading "./" as meaning "relative to
    current dir" (similarly to what is done in most shells), or maybe
    make a C-u prefix tell load-library that the initial minibuffer
    content should be default-directory.  Then maybe we could imagine
    getting rid of load-file.

I almost wrote something similar (the "./" part) in my previous message, but
I think that 1) it would be confusing and somewhat hidden functionality, and
2) people would also miss "../", "~/" etc. - IOW, they would miss completion
from the `default-directory'.

Yes, a C-u could signal a change from the completing behavior of
`load-library' (absolute file name) to that of `load-file' (relative file
name, default-directory). But why do that? What is the advantage of getting
rid of `load-file', with its handy UI for file-name completion?