Re: Multi-sessioning for GNOME [revisited]

[Havoc Pennington <hp redhat com>]

Let's move this to gconf-list -

Colm Smyth <Colm Smyth ireland sun com> writes: 
> Adding hints to configuration variables is in my view a bad idea:

What I mean by "hints to configuration variables" is that apps would
be able to affect the partition map. i.e. we would have a partition
alias "display-dependent" and apps could say "if the user hasn't said
otherwise, this directory/key should be mapped to the
'display-dependent' partition, if it exists".

> - requires programming effort in every application

It should be just a file of some kind, not programmatic. Similar to
the .schemas files. Or it could just be the same file, even though
this isn't part of the schema.

> - creates configuration data that is forked along fixed meta-variables
>   (such as display)

That is true.

> - complicates back-end storage

I don't think this is true (the hints are not in the backend, they are
just an issue of which backend gconfd will use).
 
> By separating different sub-configurations into different
> GConf partitions, apps don't have to maintain 'hints'
> for individual configuration keys and the ways of partitioning
> and combining different parts of each user's desktop's total
> configuration are infinite.

I think I agree here (I agree that we should have a generic partition
map; I'm not totally convinced of all implementation details,
e.g. where the map is stored and how, etc.). But I don't think this is 
in conflict with apps providing a default partition map for their
keys, and GNOME shipping with some useful partitions such as
"display-dependent" already set up. That way things work properly out
of the box. Otherwise the user has to somehow go through the GConf
database and find all the keys they want to be specific to a display
and check them off or something.

Colm we should get on the phone sometime and hash out GConf 2.0
plans. ;-) Maybe not immediately, but before we start working on it in
earnest.

Havoc