Another Journal Ideas

[C. Scott Ananian]

Yes, what you've described is more-or-less the plan of record: don't
store any metadata which can be extracted from the actual content, and
use plugins in the indexing service to extract interesting metadata
from a variety of "real" formats.  The few bits of metadata which
can't be representing in existing document formats (non-path tags,
"action id", etc) will be stored in xattrs, which are compatible with
all modern Unix systems (including Mac OS X).  They may be lost during
transport to non-XOs, which is why we'll try to minimize the number of
these bits as much as possible.  This also makes us as robust as
possible against indexer failures: we should always be able to rebuild
the index using nothing but the information on disk (and possibly
taking advantage of better indexers and metadata extractors when we do
so).

Non-local search is possible (see part 4 of my journal screencast),
but probably not transparently: you will have to select a particular
friend whose journal you want to look in.  This seems more reasonable
(to me, at least) in an environment without perfect connectivity: if
you try to search everyone's journal all the time, it's hard to know
whether you didn't find any matches for your search because there
really weren't any, or if the person who had the matching document
wasn't on line.  If you explicitly selected the person to search, it
can be made a lot more obvious whether or not their journal is
available to search at this moment in time. (Where "available to
search" may mean "cached on a local schoolserver"; it doesn't
necessarily have to mean, "my friend's XO is turned on and connected
to the internet right now".)
 --scott

-- 
                         ( http://cscott.net/ )