On 21/01/2005, at 5:40 PM, Tarun Ramakrishna wrote:
Lovely debate! :-)
I would vote for 1.5! The java.util.concurrent package is brilliant and could be utilized. There's also the Scanner class!. :-) . Developers who want to use the old JDK's should use the current or older versions. If the new language features and libraries can make life easier it makes sense to use 1.5 for the latest and greatest.
This has been mentioned a few times now - people who are using old versions of the JVM do so for a huge range of reasons, often because changing would require them giving up very big deals as customers aren't willing to upgrade and go through the hassle of enterprise wide deployments of a new JVM.
Despite this, HttpClient users will still need fixes and would most likely appreciate new features and design improvements. HttpClient needs to remember that it is a library and it's useless unless it is easy to integrate into an application - if HttpClient forces an application to drop support for a JRE that's a serious headache.
I'm a very, very, very strong -1 (non-veto only because I think vetoing would be pretty rude in this case) on requiring Java 1.5. Doing so would mean dropping Mac support and most likely permanently dropping support for Mac OS X 10.3 and below.
I'm -1 (again non-veto) to requiring Java 1.4 when it is possible to avoid it. NIO may make it worth moving to Java 1.4 but I'm not convinced of that. Moving to Java 1.4 support means dropping support for any applets that run in Mozilla (and variants) on OS X and permanently dropping support for applets in IE on OS X. You'd also be dropping support for most deployed J2EE servers which tend to be running Java 1.3 (though 1.4 is starting to gain momentum server side).
So I'd be keen to support 1.3 if we can because it's about the oldest JVM (apart from the MS JVM 1.1s that are everywhere) that applications are likely to commonly run into.
I know that new language features are always cool and it can be painful to have to work around missing APIs or conditionally implement functionality etc, but dropping support for JVMs has a huge knock-on effect which is much bigger than most people first realize. We need to keep in mind that HttpClient is a library that has to fit in with other applications requirements and demand as little as possible. As such I'd really like to focus on the functionality we want to provide and then look at what JVM is required to deliver that functionality.
Also realize that if we leave a lot of users behind on an older version we will need to continue to support that version by fixing bugs etc in it for a long period - maintaining two code branches may prove more time consuming than putting up with the limitations of an earlier JVM.
Tarun
Regards,
Adrian "I Still Have To Work With Java 1.3" Sutton ---------------------------------------------- Intencha "tomorrow's technology today" Ph: 3420 4584 0422236329 35 Prenzler St Upper Mount Gravatt 4122 Australia QLD www.intencha.com
PGP.sig
Description: This is a digitally signed message part
