Having a separate implementation makes sense to me.
We still need the SCA for any contributions to the FI project. The way
it works, the contributor keeps the copyright assignment, but Sun also
has the copyright.
Paul Sandoz wrote:
> Eduardo Pelegri-Llopart wrote:
>
>> Now, back to packages. We are using com.sun.* because that is what we
>> use directly inside our implementation. I'd use the same for your
>> FIME pieces.
>>
>
> Right, i think it best to use what implementation specific packages make
> sense to you.
>
> We still have to work out dependencies between FISE and FIME code. I
> think this may take a little time. My preference would be to do a
> complete new implementation of FIME without impl dependencies on FISE
> i.e. fork it. This is because the optimization strategies are completely
> different for ME and SE e.g. FISE provides a base layer for multiple
> parsers and this may not be the most optimal for a StAX only
> parser/serializer for FIME.
>
> After we can take a step back and look at what correlations we can see
> for potential sharing.
>
>
>> Are we actually using org.jvnet anywhere in the FI code today?
>>
>
> Only for interfaces and non-implementation specific code. This is still
> only half-backed. The vocabulary APIs need to be defined and i would be
> interested to know what APIs might also be suitable for FIME because
> reducing the size of messages is quite inmportant in this space, hence
> using an external vocabulary could be rather useful.
>
> Paul.
>