Re: Generalised approach to storing address details

From: Marshall <marshall.spight_at_gmail.com>
Date: 13 Dec 2006 14:45:44 -0800
Message-ID: <1166049944.448935.11020_at_80g2000cwy.googlegroups.com>


On Dec 13, 8:16 am, "Neo" <neo55..._at_hotmail.com> wrote:
> > It has been proven in numerous research studies that menu's use is deletrious
> > if nesting goes above two levels... File systems are the only real vestige of
> > hierarchy left and are a well known and are rapidly changing to add none
> > hierarchical features...
>
> If hierarchal interfaces are so deletrious and we have numerous
> research studies to prove it, why not extend this to the file explorer
> and limit it two levels?

Why not indeed?

Hierarchy in filesystems today is mostly a matter of inertia. When filesystems
were first evolving (1960s?) computer systems were quite simple and limited,
and hierarchy is easy to implement. (Lameness is often an implementation
advantage; less often is it an interface advantage.) By at the latest the 1970s
it was already apparent that a strict hierarchy was inadequate; witness the advent of hard links, and later soft links, in the Unix filesystems; this is
an implicit admission of the inadequacy of trees.

If search was better we wouldn't have directory trees. Generally the existing vendors of filesystems have little reason to innovate; hence the rise of search in web indexing, where the work is being done by the (former) upstarts who have no investment in the status quo. Also there is the difficult matter of interfaces. Existing OSs are deeply wedded
to open/seek/read/write/close. Changing the interface the OS has to the filesystem would be a painful undertaking.

> > Very little in the world is hierarchical.
> How about physicals things like cars and airplanes?

Physical things have atoms which form molecules. Any other organizational principle is an abstraction made by the mind, not an attribute of the object.

> How about B-Trees that are used to implement most relational databases?

I would agree that trees of various kinds, including B-Trees, do in fact form trees.

Trees do have their uses. They can be a good fit for when nodes have no particular relationship with each other. Good examples include the Java package system and DNS. Both namespaces are hierarchical; in both case the nodes (packages, domains) are unrelated other than by name. Abstract Syntax Trees and expression trees also fit nicely into trees.

Marshall Received on Wed Dec 13 2006 - 23:45:44 CET

Original text of this message