On Wed, 7 Nov 2001, Tim Rowledge wrote:

> While we're worrying about filenames, how about adopting the technique
> we settled upon at Interval; all filenames are URLs, and only the file
> handling prim code ever sees anything else. People are getting used to
> URLs, they have reasonably documented rules, they open the door to
> a nice interace to networks, all sorts of advantages.

How do file URLs handle relative file names?  I appreciate Squeak's notion
of relative file names because it matches my own style of programming.  I
want to create / manipulate files relative to something (in Squeak's case,
the location of the image/changes files).

If Squeak were to use URLs, I'd like to use them in some consistent way.
There is already some support for this in the HTTPClient architecture, but
it would be neat to do something like

http://my.server.com/pretty.gif (get a picture or some resource)
ftp://my.server.com/pretty.gif (ditto)
superswiki://bobs.superswiki.net/myCoolProject.pr (get the latest project
from the SuperSwiki)

All of these things exist, but all over the place.  Some consistent
support for this would be nice.

Perhaps a hierarchy like
  Url handling
    ftp handling
    http handling
      WebDAV handling
      XMLRPC handling
      Project server handling (like the existing ServerDirectory stuff)
        SuperSwiki projects
    file handling
    nebraska handling (?)
    other eToy peer-to-peer handling (?)

I am thinking aloud, so to speak.  But I feel like Squeak has a lot of
filesystem-aware and network-aware features, and I don't know how to
organize this space.

-Eric