RFR(m): 8139233 add initial compact immutable collection implementations

Stuart Marks stuart.marks at oracle.com
Thu May 5 21:00:02 UTC 2016



On 5/5/16 6:42 AM, Chris Hegarty wrote:
> I have some sympathy for this. Since we are specifying that these
> Collections are serializable, it would be nice for them to deserialize
> on an older release, but I don’t think we should compromise the
> serial form to achieve that.  In fact, I agree with Stephen, I think we
> should go the same way as that of the serial form of 310 classes.
>
> So is there anything we can do for serial interoperability? What if,
> we agree the serial form and the private supporting implementation
> types in 9 ( say, some time this month ). Backport these to, at least,
> 8u, where they do not impact spec and are benign. JDK 9 is not
> scheduled to ship until March 2017, with the regular Critical Patch
> Updates that most vendors provide, I’d bet that a large number of
> 8u’s in production at that time would contain the required types to
> deserialize these collections.   There’s no guarantee of course, but
> in practical terms at least we made a reasonable effort.

Interesting approach.

One could backport just the serial proxy and then have its readResolve() 
instantiate its existing collections wrapped with unmodifiable wrappers. Or, one 
could just backport the immutable implementations. There's not much code to 
backport.

I guess the question is whether this is worth it. Is this something we know will 
happen, or would it be done "just in case" ?

s'marks



More information about the core-libs-dev mailing list