Re: Wim's Conf Proposal

On 15 Feb 2005 at 13:06, alex black wrote:

> hi all,
> 
> I like wim's conf proposal. Some points:
> 
> 	-I would very much like to see no XML. Database would be _much_ better.

Yes, it is better. I do have a subclass for reading xml, ini, php though.

> 	-The basic path-keys are very good. They would be easy to catalog and 
> they impose a little logical structure on the values
> 	-We could use damn-near the same code for string IDs if we chose to 
> use them (more on that in a minute)

Ndf's as well. Phingless.

> 	-I propose to kill aliases until such time as we deem them necessary 
> to the core.

Ok. It does confuse. ;-(

> 	-I propose to kill the method "change()" and keep "set()".

Ok. The method change was ment for persistency, btw. It did sneak in 
sometime.

> 	-I propose to kill the method "locate()" which seems to be 
> file-specific.?

Yes, it is file specific.

> 	-I propose to kill the method "value()" since it does the same thing 
> as get()?

Ok. In case the conf_value is an array, 'value' would give you a var_dump 
or so; 'get' would give the (multilevel) array.

> 	-I propose to have set() throw exceptions if there are problems 
> connecting or the value is bad.

Ok.

> 	-I like manuel's idea to have 'levels' in a single table. (factory, 
> workspace, user)

Don't know how that would fit in the Api, without an extra parameter.
What about 'scope' ? Replace that with level?

> 		-Do people believe "user" is a good idea? (We could allow conf to be 
> used in runtime systems for pref storage - it's a perfect dictionary 
> storage thing that could use SimpleEntity.
> 	-I propose to have _no_ preloading of configuration values - they 
> should be fetched and cached by conf only when requested.

No backward- or forward- compatibilty with current conf;?


> So basically I am proposing that conf use a 'dictionary' simpleEntity 
> with very basic read/write methods that include simple validation.
> 
> I would like to see Locale use this same dictionary simpleEntity for 
> translation services using string keys that look the same as conf. We 
> _could_ allow aliases in the dictionary entity so that string keys 
> could reference other string keys.
> 
> Do people agree with this?

Agreed. 
Issue: scope? or level?
issue: compatiblity or change source bc_3 ?

I'll wait for all the comments. Thursday I'll hack the class-code and test 
it.

Cheers,
wim niemans