Re: dbus and GNOME 2.8

[Curtis Hovey <sinzui cox net>]

On Mon, 2004-04-05 at 19:24, Robert Love wrote:
> On Mon, 2004-04-05 at 19:11, Rob Adams wrote:
> > So should gconf be deprecated in favor of Storage?  Or perhaps
> > reimplemented using a Storage backend?  (Long term, obviously)
> 
> No, I don't think so.
> 
> > It would certainly be cool to be able to do things like type into a box
> > "launch the application I set to have a blue background yesterday
> > afternoon".
> 
> Gconf is well suited to key/value pairs as used in application
> properties.
> 
> Using Storage to store strongly-typed well-formed data is what Storage
> would be good for.  See WinFS.  It makes a lot of sense to use Storage
> (or something similar, again see WinFS) to store data such as address
> book entries.

Well that is not exactly true.  It doesn't have a schema concept to
enforce well-formedness and validity.  Moreover, Storage cannot deal
with the rich data-types commonly found in XML markup; there are issues
storing, querying, and retrieving data.  The matter is complicated by
Storage's goal of holding irregular XML

These issues are the same issue facing all developers and companies
struggling to fit rich heirarchal data into sparse simple tables.  A lot
was achieved when Cobb introduced relational database and SQL in the
areas of performance and scalability.  XML databases have yet to achieve
a modicum performance and scalability.

Storage is the right long term solution.

> What the goal of Storage, WinFS, etc. is is to decouple the data from
> the structure of the file.  Right now, complex data formats essentially
> recreate file system structure inside of each file.  The filesystem
> should be the ultimate storage medium.  To realize this, and have
> everything-as-a-file and one-thing-per-file, we need a much smarter
> filesystem.  WinFS can natively understand complex concepts such as
> "address book entries" in a very simple and clean way, because the file
> system itself implements structure and typing.

DBs do their own filesystem management, supporting block size, type, and
index mechanisms for performance.  The plethora of modern filesystems
fail to address the meaning/query problems that users and applications
need to deal with.  Just having a good metadata db to query and make
sense of the filesystem goes a long way.

-- 
__C U R T I S  C.  H O V E Y____________________
sinzui cox net
Guilty of stealing everything I am.