[sci.math] Formalizing location and distance in file retrieval systems

efam@nvuxr (elliott familant) (08/07/90)

I am trying to develop a descriptive framework to characterize different
kinds of file operating systems that might be found on a computer.  I am
specifically interested in formalizing what is meant by location and
distance in various file systems.  The two kinds of file systems around
today are hierchical retrieval systems (found in UNIX, DOS, VMS,etc.)
and those based on a desk top metaphor.  Distance, in the formal case,
corresponds, I assume, to some graph theorectical notion of distance
(since a hierarchy is simply a tree structure).  Distance in the latter
case, is more like what it's like in real life.  The two kinds of
distances differ, but how?  Is it correct to say that distance on a
graph is measured on an ordinal scale, whereas distance on a desktop is
measured on a ratio scale?

And what of location?  Location on a "desktop" interface is kind of
strange.  It does not correspond to a single abstract point in space,
but represents a set of points.  How can I talk about this kind of
location in some kind of formal language. 

Please forgive the simpleness of my questions.  My training is not in
mathematics so if my questions sound naive, they probably are.  But if
anyone can make suggestions as to ways to think about this, or can at
least direct me to some body of mathematics that would be relevant to
these kinds of issues, I would be greatful.  

Please send me Email if you have any thoughts about this.  

Thanks in advance.

Elliott Familant

pitchers@prlhp1.prl.philips.co.uk (Steve Pitchers) (08/08/90)

Actually, the Unix filesystem is more than just a hierarchical tree.
Any file can be linked to any other, using either a hard link or a soft
link - see 'man ln'.

A file may belong to more than one directory, which would really cause 
havoc with the notion of 'distance'.  The 'distance' to a file
will vary depending on the route you take.

The super-user can even forge multiple hard links to directories -
completely breaking the illusion of a hierarchical tree.  The resulting 
potentional for confusion explians why this is restricted to being a
super-user privilege! 

On the positive side, the existence of alternative routes to files
can make it much easier to locate information than if a 'hierarchical
menu' approach has been used.  Taking the 'wrong path' can
still lead to the correct information, as alternative opinions on the
classification of the information can be catered for.


Steve   ---   pitchers@prl.philips.co.uk
   ~~~

bjornl@sics.se (Bj|rn Lisper) (08/09/90)

In article <1129@prlhp1.prl.philips.co.uk> pitchers@prlhp1.prl.philips.co.uk
(Steve Pitchers) writes:

%Actually, the Unix filesystem is more than just a hierarchical tree.
%Any file can be linked to any other, using either a hard link or a soft
%link - see 'man ln'.

%A file may belong to more than one directory, which would really cause 
%havoc with the notion of 'distance'.  The 'distance' to a file
%will vary depending on the route you take.

You could simply define the distance to be the length of the *shortest*
path.

Bjorn Lisper

jim@se-sd.SanDiego.NCR.COM (Jim Ruehlin, Cognitologist domesticus) (08/11/90)

In article <1990Aug9.094718.5021@sics.se> bjornl@sics.se (Bj|rn Lisper) writes:
>%A file may belong to more than one directory, which would really cause 
>%havoc with the notion of 'distance'.  The 'distance' to a file
>%will vary depending on the route you take.
>
>You could simply define the distance to be the length of the *shortest*
>path.

This could be a subjective issue, depending on the user.  To some, distance
could mean conceptual distance, where a deeply nested subdirectory is
considered closer to it's ancestor than it's ancestor is to its sister.

- Jim Ruehlin

neff@ibmpa (randall b neff) (08/14/90)

I have always thought of directories as being containers where you put 
files based on a semantic grouping; similar to how an outline structures 
a document.  So files that "go together" get put in the same directory.
Also directories are frequently used for versioning.   The desktop metaphor
of folders models the same notion of collecting together those files that
"belong" together.  Symbolic and hard links both allow placing a file in two
or more places that it "belongs".   Symbolic links also allow hiding a 
physical structure (because of hard disk partitions or nfs mounts) under the
user's semantic structuring.