I have some stuff I did on the plane from Summit. I'm going to go back over it 
tomorrow and see if I included everything I wanted. I'll send the diff to the 
list. 

Sent from my iPhone

On Jul 1, 2012, at 9:43, Boleslaw Dawidowicz <[email protected]> 
wrote:

> I think it all comes down (again) to use cases [0] which we think this API 
> should address. Current ones are mainly around typical web application 
> development and proposed API prototype was focusing on easy usage. My main 
> fear is to go end up with too abstract and generic design that won't be 
> really appealing to most of developers. If you look at use cases addressed by 
> JSR 351 [1] it is all about dealing with attributes and scenarios closer to 
> healthcare - and this so far results in quite different API design. API which 
> I believe will be less appealing to typical web developer who doesn't care 
> about such use cases.  
> 
> More inline below
> 
> [0] 
> https://cwiki.apache.org/confluence/display/DeltaSpike/Security+Module+Drafts
> [1] http://java.net/projects/identity-api-spec/pages/UseCases
> 
> On Jun 29, 2012, at 4:56 PM, Jason Porter wrote:
> 
>> == Security IDM prototype feedback ==
>> 
>> * Fold PermissionQuery into PermissionManager to give the developers a
>> similar model as JPA.
>> 
>> * Because a Group is an IdentityType, I wonder if it would be better to
>> just have one set of methods and pass in the IdentityType sub interface you
>> would like to receive back
> 
> I guess thats a bit matter of taste. We could prepare quick prototype and 
> compare which one looks more useful to majority of folks here.
> 
>> 
>> * I feel the methods on UserQuery are restricting how the IDM can be used.
>> We can't always guarantee that there will be a first name, last name,
>> email, etc. However, being completely open and free form like what was done
>> in the PicketLink IDM is too much. Perhaps we can find some sort of balance
>> or meta model similar to JPA?
>> 
>> * Maybe we could have a base no method interface for things and fold all
>> the Query and Manager objects into one (similar to above about JPA)
>> 
>> * We probably don't want to go down the route of create a query language,
>> but we almost have a DSL with a fluent API here. Maybe we should think
>> about about JPA the names out a bit more and actually create a fluent API
>> for the Query objects
> 
> Would you have time to prepare some simple prototype like this to discuss? 
> I'm not sure how much generic design are you proposing in fact. Just some 
> bare code snippets would do to back your suggestions with some example usage.
> 
> UserQuery is not something I'm perfectly satisfied with however it is still 
> an attempt to have something similar to JPA Criteria Query - which is very 
> intuitive and easy to use IMO. I would love to see more feedback from others 
> on how to improve this part in fact. 
> 
>> 
>> * I believe we should remove some of the exception constructors and make a
>> message required (this would probably be good to have in all of
>> DeltaSpike). Adding a cause is great, but being able to create an exception
>> without a message is less helpful for everyone.
> 
> Good point. I think this part is simply not really polished yet.
> 
>> 
>> * Is there a point for having addObject and addObjects type methods? Would
>> one that's simply a varag be enough? We could check the type that comes in
>> if thy send us a collection, or simply have varags and collection
>> addObjects methods.
> 
> Sounds good to me. Again this is probably more matter of taste and 
> consistency with other APIs in the project. Maybe there should be some 
> general project wide design guideline for choices like this. 
> 
> 
>> 
>> 
>> -- 
>> Jason Porter
>> http://lightguard-jp.blogspot.com
>> http://twitter.com/lightguardjp
>> 
>> Software Engineer
>> Open Source Advocate
>> Author of Seam Catch - Next Generation Java Exception Handling
>> 
>> PGP key id: 926CCFF5
>> PGP key available at: keyserver.net, pgp.mit.edu
> 

Reply via email to