If this tool really works then it would be preferable not to backport everything to 1.4. If it almost-but-not-quite works it could be a big headache. How much experience do folks have with this?
On the MINA list this was discussed and IIRC at least one person on that list had positive experiences with retrotranslator (although it could have been retroweaver...) It is generally run at build time so there should not be any run-time performance hit - although even if there was presumably anyone using 1.4 isn't overly concerned with performance? Since 1.6 is going to be GA'd very soon, by supporting 1.5 and 1.6 we would be supporting the two most recent GA'd JDK versions. Originally we wanted to support 1.4 and 1.5 on the client since we were following the "most recent two JDK versions" rule at that time (which was something like 18 months ago). The main advantage I see of making the client codebase depend on 1.5 would be that some more code could be pushed into common. Anyone looking at the code will spot that there is quite a lot of similarity and to a certain extent duplication between some classes in the client and the broker, with the broker using generics. Marnie, can you elaborate on the 1.4 requirement? Do we have any (JPMC) people asking for 1.4? RG
