[BBB-DEV] Re: Re: Re: Re: org.jdesktop.beansbinding.Property should be interface

[Witold Szczerba <pljosh.mail@gmail.com>]

2009/7/15 Fabrizio Giudici <fabrizio.giudici@tidalwave.it>:
> I understand your point, but breaking binary compatibility is not a simple
> matter - for instance, BBB wants to be an implementation of JSR-295, and as
> such it must binary compatible with it. In any case, retrocompatibility was
> one of the key points of BBB since when it was introduced and I don't think
> it should be changed now. At least, we should make a deeper investigation
> about the impacts of the move.

What does it mean to break _binary_ compatibility with a JSR-295 which
is in fact, a document describing some idea about how should beans
binding work? Baked by dead BeansBinding project... How many times was
it suspended? Three already?
The other thing is, the JSR-295 was one of these specification
requests which was invented first, then (hardly) implemented which
means even if the document say something it doesn't neccesarly mean it
does have sense or is proven. I mean - if something is wrong with
JSR-295 I think we should rather change that instead of trying to
drive around problems.
More than that, in this very case, if there will be some extra
interface like "Prop" and then other parties will use that, following
best practices to declare vars/fields using most common supertype,
just like:
List x = new ArrayList() instead of
AbstractList x = new ArrayList()
, and there is no a single word about "Prop" in JSR-295 - then
software based on BBB will be unable to switch to another
implementation which in practice is even worse than breaking binary
compatibility. In the abstract class Property into interface Property
transition most of the software will compile, while the extra Prop
interface will destroy compatibility in hundreds places in an average
application.

Regards,
Witold Szczerba