Re: [xml-dev] The illusion of simplicity and low cost in datadesign and

["C. M. Sperberg-McQueen" <cmsmcq@blackmesatech.com>]

"Liam R. E. Quin" <liam@fromoldbooks.org> writes:

> The trouble with the resource/data fork model is that it's fragile: the
> two halves can easily get separated.  The same is true for invisible
> metadata stored as file system attributes (e.g. in Linux ext3 or ext4).

True - and true in other contexts, too.

  - The internal metadata in a book (title page, authorship attribution,
    copyright information, table of contents) may be inaccurate
    (e.g. when printers or authors were sailing close to the wind,
    legally, and wished not to be identified), but there are pretty good
    reasons that book publishers in Europe started including title pages
    in their books, not too long after the introduction of printing with
    movable type ("not too long", meaning I don't know exactly when,
    although I think it was within fifty years).

  - I was told long ago that one reason data repositories distributed
    social-science data in Osiris format, even though almost no users
    actually used Osiris to work with the data, was that an Osiris file
    always consisted of one part with the data and another part with the
    codebook (metadata), so if files were exchanged using Osiris format
    the metadata would not get lost on the way.

> So there's a lot to be said for the Unix "magic number" approach, where
> files start with a string that identifies them - "#! /bin/sh" for
> example :) - because it is carried along with the file.

Yes. Although I remember resenting having to start a Rexx program that
way on a Unix system, since in fact it's not legal Rexx.  (I expect it
feels less intrusive if you assume that everyone in the world uses '# to
the end of the line' as a comment delimiter -- or at least: everyone who
matters.)  It meant that Rexx programs I could run on one system I used
had to be changed before they could run on other systems I used.

> Maybe i'm just wedded to the creative possibilities of systems that
> encourage us to combine tools.

Hear, hear.

-- 
C. M. Sperberg-McQueen
Black Mesa Technologies LLC
http://blackmesatech.com