On Wed, 9 Jan 2002, Paul Hammant wrote:
> Date: Wed, 09 Jan 2002 12:55:14 +0000
> From: Paul Hammant <[EMAIL PROTECTED]>
> Reply-To: Tomcat Developers List <[EMAIL PROTECTED]>
> To: Tomcat Developers List <[EMAIL PROTECTED]>
> Subject: Re: Todo list for 4.0.2 b2
>
> Remy,
>
> >If our helper object for embedding doesn't fit your needs, I suggest you
> >write your own instead. It doesn't take long, and it will do what you want
> >(including having an interface, so it fits into your virtual OS dream).
> >
> I am talking about Avalon (another Jakarta project). It is not a
> 'dream' We already have a third-party web server running in it -
> Hendrik Schreiber's most excellent 'Jo!'. It is quite embarrasing that
> we do not yet (but are so close) have Catalina running in it.
>
> >There are real world users of the current Embedded class (including JBoss
> >and the J2EE RI), and none of them would see any benefit about having an
> >interface instead of object. Since I'd like to avoid unseless changes for
> >the users and also keep the design simpler if I can, I will NOT change the
> >current Embedded object to do what you propose, and I vote -1 to the
> >associated patch.
> >
> I am going ask ask you again, *Please* consider the merit for this, It
> is not unreasonable. The change is not useless as you say. I was not
> incorrect suggesting another reason for interface/impl separation (it
> was not new to you dude, it had just slipped your mind) was I?. It is
> such a small thing to do. Please dude.....
>
Paul, this needs to be turned around as well. *Please* consider that what
you are asking for has a very substantial backwards compatibility cost --
making this change would mean it's impossible to have compiled Java code
that works with both 4.0.1 and 4.0.2, because the class inheritance of the
Embedded class would change. We just went through this trying to undo the
poor factoring of org.apache.net.ServerSocketFactory -- in the end, we had
to undo an obviously beneficial change for this reason.
It is not appropriate to break binary compatibility in the 4.0 branch.
Breaking it between 4.0 and 4.1-dev is not necessarily as absolute, but
the penalty is still severe.
> Regards,
>
> - Paul H
>
Craig
--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
