Skip to main content

[JSR-354] Re: JSR 354 Outlook

  • From: Werner Keil <werner.keil@...>
  • To: jcurrency_mail@...
  • Subject: [JSR-354] Re: JSR 354 Outlook
  • Date: Sat, 18 May 2013 20:46:30 +0200


Thanks for wrapping up what happened in Zurich and the ideas below.

As for the first one Final (closed sounds a bit odd, it includes stopped, dormant, etc. and we don't aim for any of these)

Regarding full embracement of Java 8, it is still very much in flux, especially 310 and other non Final specs (we heard, the RI is done there without taking care of the spec so far) I guess we need to speak with interested parties like Google about their suggestion and see, how e.g. GWT uses JodaTime on similar occasions.
Using 310 internally and aiming for Java 8 in the RI seems perfectly fine, though it means, we won't go final before mid 2014, given Java 8 and everything in it will not be final before that either. Since the November estimate for 1.0 sounded a bit bold, maybe a risk worth taking.
If any part of the spec, especially core parts like MonetaryAmount or CurrencyUnit (convert already uses BigDecimal, there tie-ins could be easier to stand) used a 310 type, it makes it useless not just to Embedded, but the entire EE 6/7 market share.
I highly recommend asking the players (e.g. those in the EC like IBM, JBoss, SAP,...) and of course EE Spec Leads Bill and Linda about that, too.

A "backport" like this threeten-bp is just a pet project and even those in the industry like Google using JodaTime internally are unlikely to use it widely.
Imagine CurrencyUnit has some method like
isValidFrom(Instant) what would the backport use?
Calendar, Date, long or Long? That's a risk of gross incompatibility. And remember, 310 is not THE Date/Time API in SE8, it is just ONE of THREE (or 4 if you count java.util.concurrent and its TimeUnit;-) and even Date/Calendar get significant overhaul in Java 8, Oracle wouldn't waste effort if it was pruned in 9 or 10;-)
Calendar gets toInstant() hooks for the new 310 approach, unless there is a reverse method in 310, using it in the spec/API would breach compatibility with EE7 code, and even that's only out now.
So ask EE folks, starting with your colleagues at JAP, if they're all willing to accept a backport (not '' there must be only one spec, so it'll be or similar;-) until EE8 is adopted by all these platforms and products, if they do, I'd be OK with it.

More later, I'm low on battery in Malmø


Am 18.05.2013 17:43 schrieb "Anatole Tresch" <atsticks@...>:

[JSR-354] Re: JSR 354 Outlook

Werner Keil 05/18/2013
Please Confirm