On Mon, Apr 18, 2011 at 1:35 PM, Daniel F. Savarese <[email protected]> wrote: > > I guess I had more to say, or rather ask. > > In message <[email protected]>, sebb writes: >>really necessary, because of the additional work that it causes all >>downstream users. > > What additional work? As far as I know, end users--as in people who don't > write code--don't download new versions of Commons Net jars and plug them > into the applications that they use. They don't even know the applications > they use depend on Commons Net. Developers of the applications they use > update the jars when they release a new version of their software and > deliver it to their users. Developers are our downstream users and they > compile their code before releasing it. So as long as we remain > compile-time compatible, there's no problem. What am I missing? That's > not a rhetorical question. > > I must be missing some use case you have in mind, such as something > analogous to a Linux distribution updating /usr/lib/libstdc++.so or some > other shared library and breaking all dynamically linked user-compiled > binaries dependent on it. The only Java examples of that I can think of > are the result of lack of care in deploying applications (where you > end up pulling jars into your CLASSPATH that you don't intend).
My understanding is that we want to protect people who update without doing any testing (or recompiling). Personally I agree with you, our stance on binary compatibility is entirely too dogmatic and leads to stagnant outdated codebases. Hen --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
