very simple datastore reimplementation

[Jameson "Chema" Quinn]

>
> -If UID can hide in the metadata, which, if I understand, is preserved as
> part of the file even on foreign (unix-only?) filesystems (wow!), I do not
> see any compelling reason for it to be in the filename. Locally-stored files
> could have their real filenames, with 2 random characters at a time added in
> case of collisions; for external storage, collisions could simply be
> forbidden, and you could rename or choose a different directory.


I re-read the OLPFS page on the wiki and realized that, of course, metadata
needs to be kept separately on external VFAT devices. I also realized that,
even if all the metadata is in the index, it is scattered all over - you
need an explicit copy for efficiency. I think that this copy should be, by
default, on internal flash; that the external metadata copy in a hidden
directory (separate per-file) should be used only on import (that is, when
opening a file off external storage causes importing it). This could be the
same mechanism as the metadata-synching plugins I envisioned - whenever
adding something new to the journal, metadata would be imported, and after
that, any metadata changes from the journal would be copied to external
backup.

Jameson
ps. yes, I realize that this thread has morphed from a discussion of tomeu's
thing to OLPCFS, but I am leaving the subject for reference)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.laptop.org/pipermail/devel/attachments/20080509/a00a7428/attachment.html>