commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rahul Akolkar <>
Subject Re: [chain] Serializability of base impls
Date Thu, 01 Dec 2005 06:36:18 GMT
On 11/30/05, Niall Pemberton <> wrote:
> On 11/30/05, Rahul Akolkar <> wrote:
> > Recently, while looking at a codebase that uses Commons Chain, I was
> > drawn to the top commit message for ContextBase [1].
> >
> > I'm interested in knowing what the developer (and user too) opinions
> > are on this topic. Has anybody made any progress on this?
> >
> > Does anyone have a preference among -
> >
> >  * Keep base impls inherently serializable, but let subclass
> > implementors make their own decisions about serializability. This is
> > probably a good idea for any Commons library, since interfaces can't
> > be "un-inherited" (quoting the commit message). I think this action
> > will warrant a major release, and rework on ContextBase (to not
> > inherit HashMap) amongst other things.
> Whats wrong with HashMap - it implements Serializable?

Yes, and the subsequent implication that has on all subclasses. This
already affects (Servlet / Portlet / Faces) WebContext classes in
[chain]; but more importantly it can also affect users of Commons
Chain the same way.

"Affect" -- that is, only if you believe that classes should advertise
serializability if and *only if* they are serializable. Usage wise,
the argument can be easily made that if one is never going to attempt
to serialize a class instance, then it really doesn't matter even if
its advertised as serializable and isn't (because you'll never know
any better). For that school of thought, I had presented the third
bullet "Do nothing at all".


> Niall

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message