geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Richard Monson-Haefel <Rich...@Monson-Haefel.com>
Subject Re: Axioms for JCA implementation
Date Tue, 26 Aug 2003 05:53:23 GMT
On 8/25/03 3:51 PM, in article BAY8-DAV16HtMFaukyC00043d86@hotmail.com,
"kamesh kompella" 
<kompella_geronimo@hotmail.com> wrote:

> Hi Richard,
>                   Thanks for the info. In that case, I will look into getting
> RA going for openjms. This will be the "dummy" RA that can be used for testing
> the app server's contract's implementation once the infrastructure is in
> place.  
>  
> I was intrigued by one of your statements:
> "Personally, I don’t think the CCI offers much value. Most Ras will have very
> specific APIs. I think we should avoid supporting it."
> Is the variability in API ok as the code will be derived at deployment time
> with the help of some tools which will hide the custom API from source code?
> Also, what happens to two tier clients ( Should we leave them out altogether
> since this is as bad as mixing DB calls in a JSP) ? Do we assume that these
> clients will talk to the App server?
>  
> Thanks.
> Kamesh

If I remember correctly ­ its been a couple of years ‹ the CCI is an
abstraction for a synch resource. The idea was to create an API that any
resource could implement, regardless of its purpose. However, in practice
each type of resource has its own API and usually its easier to work with
that, than a generic abstraction.  I think some vendors have implemented the
CCI for ERP RA adapters, but I¹m not sure about that. Personally, I don¹t
think the CCI adds much value, but I don¹t really care if someone else wants
to implement it.

Richard

Mime
View raw message