Unix relative path names vs Interlisp absolute path names

[masinter]

This is a clash of styles. Interlisp grew up with the idea originally that file space was organized around directories and files within them. <DIRNAME>FILENAME.EXT;VERSION

Later was added the possibility (rarely) of adding subdirectories
<DIRNAME>SUBDIR>SUBDIR>FILENAME.EXT;VERSION
and then the notion that files were on a server
{SERVER}<DIRNAME>SUBDIR>SUBDIR>FILENAME.EXT;VERSION
and then different kinds of servers {pele:envos:mountainview}, {ftp.parc.xerox.com} (?? my memory is kind of vague as to how TCP servers were distinguished). {DSK} was OK for Interlisp-D on D-machines because it was just for Lisp -- there was no sharing of disk space with other environments.

Now in 2020 we have lots of things that don't fit, and some short-term work-arounds that may or may not get us to where we want to go. I started this as a "Discussion" rather than an Issue, which I think should be reserved for bugs or concrete plan, and rather than an email message to help focus the threads (so if you get this as an emailed notification,' please click on the "VIew it in GItHub" link and type your response in the right box).

[masinter]

One aspect of the problem is that

Medley uses full pathnames in LOADEDFILELST and other places that tie a file to the local naming conventions.

Another aspect is the need to use full pathnames in the various *DIRECTORIES variables set in MEDLEY-INIT-VARS.

One proposal is to establish a file naming convention that every system install everything in "/usr/local/lde".
Another would be to allow relative path names to a base established at runtime by an environment variable, e.g. a kind of relative device so that {MEDLEY}/lispusers could/would be used instead of {DSK}<users>larry>ilisp>medley>lispusers> everywhere

let's try to be clear about the problem we're solving as we evaluate proposed solutions.