Re: [Gnu-arch-users] Re: arch with 'special files'

[Tom Lord]

   From: Josh England <address@hidden>

   Jan Hudec wrote:
   > On Wed, Apr 06, 2005 at 13:36:14 -0700, Tom Lord wrote:
   > 
   >>   > A single 'second-order' metadata string would basically give infinite
   >>   > flexibility in terms of metadata. [...] would
   >>   > allow any number of creative facilities to be built on top of it.
   >>
   >>A unique fixed string per file would have the same effect and we
   >>already have one of those.

   I understand what you're getting at.  One could archive their own
   metadata file(s) indexed by the file ID, then manage the metadata in
   surrounding wrappers.  This actually does effectievly allow generic
   metadata to be created.  Perhaps explicitly adding it to arch is not
   necessary.  Maybe just detailing how this could work in the
   documentation would be enough.

   However, providing a fast/formal/consistent indexing method would also
   be nice, that way the metadata is always in the same place instead of
   being different in every archive. 

Yes, yes.  In extension, arch is already close to and is likely to
evolve into a fully general "tree-oriented database" with indexing and
retrieval features for fine-grained access more closely tied to
archive transactions (which themselves will expand to allow composite
transactions).  Meta-data built-in, sure sure.  At that point, and in
combination with UI frameworks like awiki and indexing tools like BDB,
Arch will pretty much make popular commercial relational database
frameworks look like crude stone tools by comparison.

But first we have to argue for 3.5 GD years about whether or not
`~/=tagging-method' is or is not a sufficiently ugly filename that it
should be replaced by an MSQL database of XML fragments using the
`my-aunties-fanny' Schema.

But I digress (it's friday),
-t