[BOB] [vfs] Abstracted

  • From: "Bas Westerbaan" <bas.westerbaan@xxxxxxxxx>
  • To: projectbob@xxxxxxxxxxxxx
  • Date: Wed, 28 Mar 2007 00:15:36 +0200

I did a bit of brainstorming today.

In my head I already had a nice concept of how a practical core vfs api
would look like and I was even busy documenting that.

Everything would be objects. A file would be an object on which you have
some operations like read and write (and others). A directory would be an
object with a list operation.

Not just the stuff in a filesystem would be objects, the core api of the
filesystem as I imagined it would be fit for being objects themselves too.

When, for instance, resolving a path to an object you are querying in a
specific manner. It shouldn't be the only way, nor should it have to be
there at all.

In other words, resolving queries is done by objects. (Directory's list
operations for hierarchical, Index objects on Directory's for keywords and
something similar for relational queries)

There is, actually now I gave it some thourough thought, nothing that _must_
be there in a filesystem except for being able to do stuff with objects and
having one starting point, a root object.

It all boils down to this: a grand unified namespace as a grand unified
filesystem would be, would be nothing more than a system-wide dynamic object

Obviously, though, this isn't the full story: we don't want to require stuff
like hierarchical querying, but we certainly do want to have a standard
implementation of it.

I'll hack something together tomorrow.

Bas Westerbaan
GPG 99BA289B | SINP bas@xxxxxxxx

Other related posts:

  • » [BOB] [vfs] Abstracted