Brian Stansberry wrote:
I'll respond to your last point first, as it's the
"bottom line."
Other responses in-line.
My impression when you first asked on this thread yesterday is you'd
decided you really did not want to disrupt your 2.6 roadmap in order
to get it in AS 5, and were seeking buy-in from Dimitris and myself.
Seems I was mistaken.
No, I said I would cut a 2.6 *before* the planned release data, with
come features pushed into 2.7, if you guys were willing to accept it
into AS 5. Of course, since this is more work for me, I would not mind
if I didn't have to do it ... :-)
If 2.6 can't interoperate with 2.5, we should get 2.6 in AS 5.
2.6 is backward compatible with 2.5 if you don't use the new connect()
call added in 2.6.
I'd think there'd be no problem upgrading across minor
releases if we
went from 5.0.x to 5.1.0, but there would likely be a problem if we
went from 5.0.0 to 5.0.1.
That is my understanding too. And if we go for 5.0/2.5 and 5.1/2.6, then
we would *not* support the 5.1/2.6 combo; according to the standard Red
Hat policy, customers would have to upgrade to 5.1 to use 2.6.
Either way, this points to a larger issue where it's unclear what
the
big picture AS roadmap is. Do we want to do a bunch of 5.0.x releases,
or fairly shortly after 5.0.0 go to a 5.1.0?
I'm interested in the answer to this too...
> Yes. But remember, this is a new method, we do *not* overload the
> existing connect().
>
>
Still confused; this and your answer to the above seem to contradict.
But no matter, I'll understand when I look at the API. :)
My yes was to "is it absolutely necessary to throw exceptions?", *NOT*
to "is upgrading to 2.6 a must", maybe that's the src of the confusion.
Key thing is 2.5 and 2.6 can't interoperate.
No, not true, they can if you don't use the new connect() methods.
--
Bela Ban
Lead JGroups / Clustering Team
JBoss - a division of Red Hat