[Zope3-dev] RFC: zcml:condition

[Martijn Faassen]

Dieter Maurer wrote:
> Martijn Faassen wrote at 2005-2-18 12:09 +0100:
> 
>>...
>>Not as powerful as before, which is good.
>  
> -1

I wasn't proposing something here, so it's rather weird that you say -1..

Presumably you are suggesting that the more power, the better? I do not 
think that is true in all cases.

>>External ZCML processors might 
>>mostly try to ignore this, though that also depends on how conflicts are 
>>dealt with -- see my question below.
> 
> In another message use gave pretty printing and validation
> as examples. I think, they could ignore it without problem...

I also listed analysis tools, which do not exist yet, but why make life 
harder on them?

The general principle is to keep the language declarative and try to 
keep it independent from external state. The more such dependencies you 
add to it the lesser the chance any static analysis tool can make sense 
of it without having to load up the full Zope 3 codebase, and I also 
think the understandability for the user also decreases.

A conditional dependent on the execution of a function defined in Python 
*will* make analysis of at least that conditional impossible witout 
requiring the full Zope 3 codebase to be loaded (as you don't know what 
it will import).

There are of course already other issues with such analysis, such as 
implements() and now adapts(), but at least static analysis has a hope 
in figuring that out in most cases.

Regards,

Martijn