> On Saturday 10 December 2011 13:35:38 Timothy Pearson wrote:
>> >> Does it drag in other KDE 4 dependencies?
>> >
>> > Obviously it depends on kdelibs and kde-runtime like any KDE
>> application.
>> > The
>> > dependencies will of course be better once frameworks 5 are released.
>> >
>> > Considering the impact on a running environment I would say that all
>> > depends
>> > whether your users are using any KDE 4 applications or none. I would
>> be
>> > surprised if your users would use no KDE 4 applications. So I think it
>> is
>> > negligible.
>>
>> Right away AFAIK we would lose DCOP integration,
> which should not be any issue for a window manager. Window managers use X
> for
> IPC as well you get D-Bus integration. Given that in the long run you want
> to
> migrate to Qt 4 and with that D-Bus anyway this sounds to me like a clear
> plus.
>> the window styles that come with TDE
> you are aware that KWin 4 ships the same window decorations (some got
> moved to
> kdeartworks) plus a theme engine? Do I have to point out that you broke
> all
> 3rd party decorations for kwin3 in such a way that kwin won't start if you
> selected one of those?
>> , and the kcontrol integration.
> Which is no issue as KWin allows to access the complete configuration from
> each window. You could even ship the old KCMs for KDE 3.5 as those have
> hardly
> changed.
>> Users can already run
>> different window managers (compiz anyone?) so I would say this decision
>> should be made on a per-user basis.  A set of instructions on the Wiki
>> for
>> configuring a users' session to use kwin4 wouldn't hurt. ;-)
> I can only recommend you to consider this offer from our side. Think about
> what you actually want for your users. What is really important to you and
> if
> you really want to develop a window manager.
>

Regardless, the process to drop/replace something is not trivial.  This
project is not like other desktops; we don't just drop or replace things
on a whim.

The process would be:
1.) Develop a procedure to replace twin with kwin4 on users' test systems
2.) Those users would need to thoroughly test the replacement for many
months, noting any regressions and having them fixed upstream.  Upstream's
response time and overall regression rate would also be evaluated at this
time.
3.) If the replacement after testing looks viable then a deployment
procedure would be created.  twin would still be available, but kwin4
would be an option, either in build or in package installation
(preferred).
4.) If the majority of end-users choose kwin4 at this point, then twin
would be deprecated but still maintained as compilable in the TDE source
tree.

As you can see this is not trivial.  These guidelines are be applied to
any integral component of TDE, including the HAL replacement, and are not
meant to be an onerous set of rules to prevent progress.  Rather they are
meant to lessen the possibility of sudden unexpected breakage for corner
case users, as has been seen many times before in open source history.

Tim