Yes and it's not just JTA that makes that kind of assumption. OTS does too, which although an older spec is still in use quite widely. As standards, expecting them to change to a new threading model such as Fibres might be nice, but it's extremely unlikely happen in any meaningful timeframe.
Absolutely agreed. And not just JTA..... I also think about:
At my most optimistic, I feel a "barrier-of-normalization" is required here. All the peculiar stuff occurs on side of the barrier, i.e. mapping fiber behavior to threads,
And on the other side there's a big sign "No Dogs, No Fibers" and it's all vanilla.
The question is, with such a barrier in place, does one still get benefit from lightweight threads.
At my most pesimistic, I imagine being on a conference call with JBoss, Oracle, and IBM trying to figure out why our 2PCs don't work, and they all yell "WTF is a fiber ?", and then I feel this is a total waste of time, but it's not like that today
But as you said, Mark, this won't happen over night.