I commented on your PR - thanks for sending that over. I think it would be
worthwhile structuring the class with the constants in such a way that the
javadoc could end up on the website via David's site generation code. That
would be extremely cool and I'm sure a very useful piece of documentation.

While I appreciate your service-jar.xml visibility comment, I suspect that
is because of something missing in its documentation, as opposed to an
issue with the design choice. I don't necessarily object to something that
can parse the config in YAML (I believe there is something that handles
JSON already), providing the existing config mechanism in XML, with the
defaults in service-jar.xml continued to work, and that I wouldn't be
forced to migrate from XML to YAML. Many users make use of tomee.xml and
resources.xml in a huge number of applications. Forcing a migration on them
to YAML "just because" doesn't make sense to me. The concept of the
service-jar.xml is a very core design concept in TomEE and I would not be
favour of changing it unless there was a really good rationale and a better
design proposed and discussed.

I agree that system properties can lead to "magic" parameters that aren't
well documented, and I'm definitely in favour of improving the
documentation of those properties. The ability to override config with
system properties, however, is useful and also widely used so we'll need to
keep it.

Jon

On Wed, Jan 2, 2019 at 9:53 AM Gurkan Erdogdu <[email protected]> wrote:

> For me, using services-jar.xml approach is not so visible to users. All
> defaults are configured in this file and packaged within JAR file. Users
> are not able to read the parameter explanation from services-jar.xml files.
> Also, services.-jar.xml is somebit different from using the system
> properties to turn-on/off something. I am also thinking to introduce YAML
> based configuration.
>
> But first step is to centralise all of these system parameters into two
> classes. Maybe, we will see that some of them are unnecessary etc.
>
> On Wed, Jan 2, 2019 at 12:44 PM Bruno Baptista <[email protected]> wrote:
>
> > Yes, there is.
> >
> > This is also the most basic MP spec and nothing prevents us from using
> > it everywhere.
> >
> > There might be Jakarta EE restrictions in how to handle configurations
> > that need to be assessed.
> >
> > Overall, I think that if we are going to mess with configs, we should
> > use state of the art.
> >
> > Cheers
> >
> > Bruno Baptista
> > https://twitter.com/brunobat_
> >
> >
> > On 02/01/19 09:35, Jean-Louis Monteiro wrote:
> > > I think with microprofile-config we may have a chicken and the egg
> > problem,
> > > isn't it?
> > > --
> > > Jean-Louis Monteiro
> > > http://twitter.com/jlouismonteiro
> > > http://www.tomitribe.com
> > >
> > >
> > > On Wed, Jan 2, 2019 at 10:30 AM Bruno Baptista <[email protected]>
> > wrote:
> > >
> > >> Hi Gurkan,
> > >>
> > >> I agree we have a problem with the documentation of the different
> > >> properties and that we need to improve it.
> > >>
> > >> Doing the inventory and using the proposed syntax looks ok to me but I
> > >> also think we should go even further.
> > >>
> > >> How about to migrate all the properties to use microprofile-config?
> > >>
> > >> Cheers.
> > >>
> > >> Bruno Baptista
> > >> https://twitter.com/brunobat_
> > >>
> > >>
> > >> On 02/01/19 07:20, Gurkan Erdogdu wrote:
> > >>> Hello
> > >>> There are lots of known and unknown system properties in the current
> > code
> > >>> base. I would like to introduce TomEESystemProperties and
> > >>> OpenEJBSystemProperties classes to hold these system property
> constants
> > >> and
> > >>> provide clear comment what it does. For example:
> > >>>
> > >>> class TomEESystemProperties{
> > >>>       public static final String TOMEE_FORCE_RELOADABLE =
> > >>> "tomee.force-reloadable";
> > >>> ....
> > >>> }
> > >>>
> > >>> class OpenEJBSystemProperties{
> > >>>      public static final String OPENEJB_CROSSCONTEXT_PROPERTY =
> > >>> "openejb.crosscontext";
> > >>> ....
> > >>> }
> > >>>
> > >>> WDYT?
> > >>> Regards.
> > >>> Gurkan
> > >>>
> >
>

Reply via email to