Re: [PROPOSAL] 3. RfD on new debian java policy

[Ben Burton <bab@debian.org>]

Hi.  Just a few more notes on this.

> Java libraries must depend on the needed working runtime
> environments, including the virtual package names of working
> versions of the "unfree" bin/java interface.

(and a similar clause for java programs)

If you're going to mandate that developers keep a list of working JVMs
for their dependencies, you should probably mandate that one of these
working JVMs be used in the startup script.  Having a JVM installed on
the system (via a depends) has no bearing on whether an app's startup
script actually calls one of these working JVMs.

Note that this is going to get messy for packagers - the packager of
each java app or library will presumably need to test this app on all
JVMs with each new release, and will also need to be aware of when each
JVM makes a new release of its own so they can test their app with the
new JVM version.

This could also create political problems for packagers who will not
install non-free software on their machines (since policy mandates that
packagers keep track of which non-free JVMs work as well).

> As "unfree" virtual maschines of a certain version almost certainly
> will run code developed for this version, all packages must also use
> kown working "unfree" interfaces, when they search for a possible
> java binary.

I would make this clearer.  I presume you mean that "when searching
for a known working JVM, non-free JVMs should be included in this search
list".  However, the text could well be misinterpreted as "the non-free
JVMs should be used wherever possible, before searching for free JVMs".

> A package should use ant to build a package ...

I disagree with this working most strongly.  According to debian policy
(section 1.1), failure to compily with a "should" directive is generally
considered a bug.  I do *not* want policy to dictate what specific
build tools a package should use.  If upstream ships with Makefiles and
the packager has made sure these Makefiles work with debian tools, it
sounds just fine to me.

I would suggest changing this paragraph to "If a package uses ant to
build the package, ...".  This section of policy should IMHO be a
simple "if-else" block, not an explicit push for developers to use ant.

Ben. :)