James Im wrote:
Just a comment about OSGi.
I've used it in the past and stopped using it (after 18 months) because
I found that in the end it added more complexity (maintaining bundles
files, when changing) and didn't than real value. I'm talking about the
chore of maintaining bundles files and a pernicious impact on my way of
coding: I was always thinking about which class could use which other
class according to OSGi wiring.
Agree. Maintaining manifests for bundles is a hassle if you do it
manually, especially if you have more than a trivial number of packages
to deal with, but the new maven-bundle-plugin alleviates much of this
maintenance hassle. In most cases you can just take the defaults and it
builds all the right metadata for you.
The kind of problems I saw happened with 35 different bundles on
different startlevels. As mina will be packaged as just one bundle you
won't have that king of problem but just as a warning I advise you not
to make any class refactoring that don't make sense, that you wouldn't
have made without OSGi.
Agree again. I think even if we didn't target OSGi the package
refactoring patterns suggested so far would make good sense.
kind regards,
John
