Re: RFC: remove print statements from Security

[Dalibor Topic]

Mark Wielaard wrote:
> Hi,
>
> On Tue, 2004-10-19 at 23:02, Bryce McKinlay wrote:
>
> > Mark Wielaard wrote:
> >
> >
> > > If you remove this could you add a little note to VM Integration Guide
> > > doc/vmintegration.texinfo about the importance of setting up the system
> > > property gnu.classpath.home.url and how it is used by Security,
> > > LogManager and InitialContext to find default configuration files. 
> >
> > Is there a reason to use a non-standard system property here? whats 
> > wrong with java.home?

[snip]

> But I am not too attached to the current implementation. And the
> gnu.classpath.home.url property is currently only used for Security,
> LogManager and the InitialContext properties. So it shouldn't be hard to
> rewrite it to something else if you care.

I'd propose to use different, better named URLs instead of a generic 
java.home one with some magic paths appended to it. I'be for using 
explicit gnu.classpath.security.configuration.url, 
gnu.classpath.logging.configuration.url and 
gnu.classpath.naming.confoguration.url properties.

The reasoning behind is that using a single magic property with some 
more path magic applied to it codifies a particular undocumented 
expectation on the source tree & installed tree layout. That's a bad 
thing in my experience.

It makes GNU Classpath unmodified harder in Kaffe, as I have to play 
tricks in the bild system, or to spend time maintaining trivial changes 
manually in sync with the GNU Classpath, as I already do for the DSO names.

cheers,
dalibor topic