Fw: Representing Unix filenames in Unicode

From: Philippe Verdy (verdy_p@wanadoo.fr)
Date: Sun Nov 27 2005 - 18:09:37 CST

  • Next message: Philippe Verdy: "Re: Representing Unix filenames in Unicode"

    From: "Hans Aberg" <haberg@math.su.se>
    > The Mac OS resource forks have been killed of as of Mac OS 10.4, now
    > being handled by the underlying UNIX in for example the case of fonts. So
    > the development is going another direction.

    Not really: the POSIX hierarchical filesystem has severe limitations, and
    work is going on to support multiple navigable dimensions in a filesystem by
    allowing as many meta-data as needed on any file, each one acting as a new
    dimension.

    The result should be a database-driven filesystem where information is
    stored according to the effective data model instead of the very limitative
    hierarchical model. This means that even the concept of "file path" would be
    remodeled in some future, by adding namespaces at each "directory level" to
    navigate from one dimension to another.

    The concept of "canonical paths" (as seen on old FAT/FAT32 filesystems)
    would disappear in favor of unique ids (or inode numbers in Unix
    filesystems), and things like "directories" would be replaced by collections
    resulting from a selection query including the navigation namespace (i.e.
    the meta-data type). Existing nightmares like "where did I put my file"
    should disappear (that's one of the reasons of the current success of local
    web search engines, that create a database of meta-data information more or
    less successfully built from the file contents...)

    Even the concept of "file" (an unqualified stream of bytes) is bad. What is
    good is the piece of information as seen in a data model. The current
    concept of file should be replaced as well as the result of a selection
    query in that model.



    This archive was generated by hypermail 2.1.5 : Tue Nov 29 2005 - 09:31:59 CST