Is there any reason left to rename packages if you're OSGi compliant? Different versions should be in different class loaders.
On Thu, May 15, 2008 at 4:59 PM, Rory Winston <[EMAIL PROTECTED]> wrote: > Just my 2c, but I think package renaming for major releases is ugly and > unecessary. > > Phil Steitz wrote: > >> On Thu, May 15, 2008 at 12:00 PM, Luc Maisonobe <[EMAIL PROTECTED]> >> wrote: >> >> >>> Phil Steitz a écrit : >>> >>> >>>> On Thu, May 15, 2008 at 1:55 AM, Emmanuel Bourg <[EMAIL PROTECTED]> >>>> wrote: >>>> >>>> >>>>> James Carman a écrit : >>>>> >>>>> >>>>>> Would it be better to make it Math 2.0? With a language level change >>>>>> like that, it should probably be a new major version. You should also >>>>>> consider putting the stuff in org.apache.commons.math2 packages to >>>>>> avoid "jar hell" issues. >>>>>> >>>>>> >>>>> And if you do so, you'll have to change the Maven identification of the >>>>> jar. >>>>> That's the right time to release Commons Math as >>>>> org.apache.commons:commons-math instead of commons-math:commons-math. >>>>> >>>>> >>>>> >>>> Both good points. Lets get the maven namespace fixed in 2.0. Thanks >>>> in advance for help making sure we don't screw that up. >>>> >>>> Regarding package renaming, there are very few backward incompatible >>>> changes lined up for 2.0, and all of the proposed 1.5-dependent stuff >>>> is new, so I am hesistant to do a wholesale change. What would be the >>>> problem with renaming only packages that contain classes with >>>> incompatible changes? >>>> >>>> >>> I see three problems with a mixed approach like this. >>> >>> - it is inconsistent and therefore difficult to understand to end-users >>> - it prevents using simple search/replace to switch from 1.x to 2.x >>> - it doesn't help when the two jars are in classpath and a class >>> with no incompatible change (i.e still in a .math.something >>> package) calls a class with an incompatible change >>> >>> >> >> Assumiing that happens, which I guess we can't rule out, so point taken. >> >> >> >>> I'm also quite reluctant to packages names change, but it is only a >>> personal >>> taste and therefore not important. I agree this is the safer solution as >>> it >>> handles things properly. >>> >>> Here is my proposal, summarizing what has been discussed so far in this >>> thread and adding my own bias: >>> >>> - commons-math 2.0 will target Java 1.5 as the minimal version >>> - the maven groupId will be changed from commons-math to >>> org.apache.commons (the artifactID is unchanged) >>> - all packages will be relocated under org.apache.commons.math2 >>> >>> If nobody complains (mainly about the third point), I am going to perform >>> these changes in branch 2.0 on Sunday. >>> >>> >> >> I guess the third point gives me heartburn. Given the relatively small >> number of incompatible changes, I do not see the need to make a >> wholesale change to the package name and force all users to make >> source-level changes to upgrade. I am +1 on fixing the maven groupID, >> though and upping the required JDK level to 1.5. >> >> Phil >> >> >>> Luc >>> >>> >>> >>>> Phil >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe, e-mail: [EMAIL PROTECTED] >>>> For additional commands, e-mail: [EMAIL PROTECTED] >>>> >>>> >>>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: [EMAIL PROTECTED] >>> For additional commands, e-mail: [EMAIL PROTECTED] >>> >>> >>> >>> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: [EMAIL PROTECTED] >> For additional commands, e-mail: [EMAIL PROTECTED] >> >> >> >> >> >> > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > >
