[IMPORTANT] NetX & Plugin Move

[Deepak Bhole]

* Thomas Fitzsimmons <fitzsim at fitzsim.org> [2010-10-19 12:21]:
> 
> 
> Hi Deepak,
> 
> Deepak Bhole <dbhole-H+wXaHxf7aLQT0dZR+AlfA at public.gmane.org> writes:
> 
> > * Dr Andrew John Hughes <ahughes-H+wXaHxf7aLQT0dZR+AlfA at public.gmane.org> [2010-10-19 07:58]:
> >> I'm working on creating the new repository for the plugin and NetX.
> >> Please don't commit any changes to NetX or the plugin until I announce
> >> the new repository.
> >
> > Nice! It will be great to have the Plugin and Web stuff out of
> > main-line.
> 
> With gcjwebplugin I went the opposite direction and merged it into GNU
> Classpath.  The advantages of that arrangement outweighed the
> inconveniences.  It meant that more people built the code regularly and
> in different environments and meant less fiddling for non-plugin
> developers wanting to test a failing applet against the development
> tree.
>

Hi Tom,

I agree with the above but with the changed landscape of the project,
the plugin-in-dev-tree model poses other new problems now which require
this move.

For one, it requires significant backporting efforts. Currently there
are 3 supported branches (1.7, 1.8, 1.9) and two tips, IcedTea6 and
IcedTea (7). Every commit has to be evaluated for these, and then
backported one by one. Once OpenJDK7 is out and IcedTea (7) becomes
generally used, it will have additional point branches.
 
Furthermore, pushing a new IcedTea/OpenJDK version requires significant testing
effort on the distro side, which limits how often it can be pushed,
which in turn limits how quickly the plugin fixes become available to 
users.

> > For one, it will avoid double commits to icedtea and
> > icedtea6 each time..
> 
> Do you foresee a need for stable and development branches of the plugin
> and NetX?
> 
> > and it will be easier for distros to update just
> > those specific parts.
> 
> Are you planning to provide standalone NetX/plugin distribution
> tarballs, separate from IcedTea?
> 

Yes to both of the above. As the plugin gains mainstream popularity,
more bugs are being found and the turnaround time for the bug fixes to
make it back to the distros is quite significant at the moment. We
definitely need to reduce that. Once the bugs become as infrequent as
for OpenJDK, we can look into folding it back. For the foreseeable
future though, I believe the split is the right thing to do.

Cheers,
Deepak

> Thomas
>