configuration system notes

[Havoc Pennington]

On Thu, 2005-04-07 at 16:42 +0100, Dave Cridland wrote:
> 
> I find this last very funny. You must realise that GConf has nearly 
> all the basic features of ACAP, indeed it's simply a cut-down ACAP in 
> order to simplify it and reduce its scope, and add strong typing and 
> schemas - that's not a bad thing, you didn't want GConf to cover the 
> same scope as ACAP can, and you didn't need the power that increases 
> its complexity.

My point is that the complexity of ACAP is not required, and the fact
that it's a standard is not interesting.

> Now, I'm absolutely not suggesting that ACAP be "the" configuration 
> system[*] - for the vast majority of desktop applications - it'd be a 
> terrible idea to have a TCP/IP protocol in use for local 
> configuration - but to dismiss it out of hand as "pretty different 
> from the ideal" strikes me as rather foolish, especially when its 
> demonstrably so close to GConf in so many ways.

ACAP specifies a bunch of coding work that isn't important, and it omits
some coding work that is important. Sure, all these configuration
systems we're talking about are "the same" on some uninteresting level,
but the details matter.

> ACAP is absolutely worth looking at as a source of inspiration for a 
> new configuration system, especially as it's the only example you 
> gave which aimed to share configuration data between wildly different 
> applications, and aimed to make it easy to manage.

I agree it's worth looking at, sure. But I don't think it's the answer
to any problem we have.

Havoc