Sylvain Wallez wrote:
So less is less when you have more on the *same* object accessed directly in Java.
Sylvain,
less is more when all the things that you removed were not helping, not always.
The "less is more" design approach is a process, not a solution.
Sure, and I totally agree with that. I should have reminded in my post the quote that's on my weblog: "perfection is achieved not when there's nothing more to add, but when there's nothing left to remove" (Antoine de Saint-Exup�ry).
Instead of putting everything in FOM and deprecate bad ideas later, we opted for a process where we start small and add thing incrementally.
Yeah, but the question is what is the FOM? Is it the objects that are made available, or the APIs on these objects? I know your answer: "both". But why do we have *two* different APIs in Cocoon for the exact same object, one being a subset of the other? This is really confusing to users.
An example: last week, a customer of mine (I do mentoring for them and they're not subscribed to users@) asked me "I get a "no such property or method" error when calling context.getRealPath() in my flowscript. Why?". I answered that this method isn't available in flowscript and provided a 10-lines workaround involving looking up the sourceresolver and resolving a "context://" URL.
Sure I could have started a vote to add getRealPath() to the FOM. But the customer needed it right now, and not in 2.1.5, and the workaround... well, just works even if ugly and slower. But it is a workaround.
It might result that we end up making FOM a java clone of the java APIs we provide. If the community requires so, great, wonderful.
That's where I see a limitation of the community dynamics and a proof that this API reduction is a bad thing, as it's faster to provide a workaround using a Java class and Cocoon's official APIs than discuss and vote some changes that will be available in the next release.
Also, the fact that the workaround uses the official Java APIs (and not clear violations of the public contract like CocoonComponentManager.getCurrentEnvironment()) clearly shows IMO that constraining the API of objects exposed by the FOM is useless.
The point is that it has been done with a process that made it appear why we needed that, instead of just cloning over.
This means: if you think there is something missing in FOM or has to change, ask for a vote to add it, at that point, you'll find resistance and people might suggest better ways to do what you had in mind, or not.
Doesn't matter the outcome, it's the community process that counts.
Ok. I'll check what has been hidden from the environment API and start a vote on this. I do agree with the fact that not all objects should be exposed (although most of them actually are), but not with the reduction of the API.
Ah, and what about the FOM if/when we'll have a Java version of flowscript? Will we have a FOMRequest Java interface that will be subset of Request? That sounds totally silly...
In *that* regard, less is always more.
Yep. Less APIs in JS than in Java for the same object is more confusion ;-)
Sylvain
-- Sylvain Wallez Anyware Technologies http://www.apache.org/~sylvain http://www.anyware-tech.com { XML, Java, Cocoon, OpenSource }*{ Training, Consulting, Projects }
