Stephen, Maybe we could call it Commons Generics? Just a thought. But, that might leave a door open to put non-collection classes in there, as generics can be any type of parameterized class. Anyway, how should we proceed from here (regardless of what we call it)? Maybe a new project in Jakarta Commons would be in order. Wouldn't this situation be analogous to the Xerces Java 2 vs. Xerces Java 1 situation? They created a new project for their "next generation." Shouldn't we? They kept the same package structure, from what I can tell (org.apache.xerces). Do you think we could do the same and let folks upgrade to the next generation without having to change their code? I don't know much about the binary compatibility issue. Would they just have to recompile their classes using a JDK5 compiler against the new classes? Of course, they'd get those new warning messages, but at least the code would work. Right?
James -----Original Message----- From: Stephen Colebourne [mailto:[EMAIL PROTECTED] Sent: Tuesday, May 24, 2005 6:56 PM To: Jakarta Commons Developers List Subject: Re: Commons-Collections enhanced with Java Generics Support John Watkinson wrote: > We are definitely ready and willing to > bring the code back in to the Jakarta Commons when the time is right. > Our goal was to get it done so that developers (us included) could start > using and improving it, but we hope that we can eventually contribute it > all back to the Commons. +1, This sounds like the right approach to me. > 1) Should it be a separate project or a branch of Commons-Collections? > > It is the case that Java 1.5 classes are NOT binary compatible with > previous runtime environments (1.4 and earlier). So, the original > Collections have a very long life ahead of them. Also, an enormous > number of third-party libraries have the Commons-Collections as a > dependency and it will not be possible to ensure binary compatibility > between the generic-enabled Commons-Collections and the original > library, even if the code is run on the 1.5 platform. There are numerous good points here. [collections] got a bad name with the 3.0 release for breaking binary compatability. (In fact this was 99% FUD spread by Hani and TSS, because only a couple of constants were incompatible). However, what it does serve to emphasise is that [collections] must not only be 'clean', but also be perceived to be clean. To me, this means that clever tools like Retroweaver would not be an option. > 2) Is there enough support for the 1.5 Collections to warrant a new > project? > > I agree with the points raised earlier on this topic. Developers may not > yet be clamoring for this, but as more of them adopt 1.5, they may begin > doing so. Also, a high-profile project like the Commons-Collections > could help attract developers to 1.5. The main problem is keeping bug fixes in sync between the two codebases. However, a separate package structure is the only practical solution to achieve 1.5 collections. Whether it is a separate project is a moot point, and I don't have a strong view either way. I would expect the package name for collections JDK1.5 to be org.apache.commons.collections15 > 3) What are the open-source credentials of Matt Hall and John Watkinson? > > We are the authors of Streamsicle (http://www.streamsicle.com). I have > also written SwarmCache (http://swarmcache.sourceforge.net) and MegaMap > (http://megamap.sourceforge.net). We are the authors of Starfish, a > project that we are in the process of opening > (http://www.larvalabs.com/starfish). Having previous OSS experience will be helpful. Finally, any ideas from http://collections15.sourceforge.net should also be taken into account. Stephen --------------------------------------------------------------------- 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]
