Re: [Summary] Meta-data/filesystem-encapsulation

[Kevin Littlejohn <darius connect com au>]

> desktop. Make the assumption even that the user's home directory is NFS
> mounted off a SunOS server. Then you will be doing the right thing.

*chuckle*  That sounds familiar....

I think we'd reached the stage where people basically agreed that writing
a library, to disassociate the yukkiness of how it's done from providing
the ability to do it to gnome apps.  (I know, there's a proper word for that,
but it's late and I'm tired :(.

I think we were also vaguely leaning toward having a system-wide db of
mime types and regexps (or hashes and regexps, or magic numbers and
regexps... whichever), but that was still a bit under dispute :)

At about that point, someone had mentioned corba, and I got really
confused, because I'll admit to still not being entirely clear as to the
consequences of implementing something like this in a corba-friendly way
(or what a corba-friendly way entails, or if it's even relevant).

There was also the problem mentioned of icons - obviously, storing each
individual icon for each individual file in a database starts to sound
very, um, registry-like :)  I still wasn't sure what the final answer
to that was - maybe (random synapse fires here) we could hold in the database
a specification as to the _name_ of the icon, and the icons could all be
held elsewhere - or maybe even hold a marker that says 'if the query is
'icon', then check this database...' - that may be a functionality that
could help us later (thinks:  If the query is 'view', and the file matches
this regexp, then check this database', where this database is an SQL thing
of some sort - dynamic creation of file contents from an SQL database?
Sounds damned nifty to me...)

Anyway, library, something messy underneath, and vague mutterings about corba
and stuff.  Oh, I was also going to point out that, if the library were
implemented nicely, it _could_ be wrapped both with little binaries (for
shell scripts), and with the LD_PRELOAD stuff, if people _really_ wanted
to do that (not that it'll solve anything outside of individual's machines,
mind, but that's my gripe... :)

Was that a fair sum-up?  I can volunteer as far as keeping a specification
for an interface, if that helps, but after all my comments, I'm afraid I
don't have the free time to actually code the thing :(:(

KevinL