On Fri, 2010-02-05 at 23:42 -0800, Adrian Crum wrote:
> Chris,
> 
> Framework independence has been a goal for quite a while. There is no
>  disagreement that the framework should run on its own. The
>  disagreements arise in what constitutes the framework.
> 
> Let's assume for a moment that framework independence means running the
>  components in the framework folder independently from anything else in
>  OFBiz. Right away the problem with that idea is that visual themes are
>  in a separate folder outside the framework folder. Does framework
>  independence include the visual themes folder? That has not been
>  discussed. Then there are the multitude of dependencies upon the
>  applications folder.

I'm a newbie here, but I have a lot of gray hair.  Seems like trying to
separate dependencies by folder or subject matter is an exercise doomed
to failure.  

TCP/IP has taken over the world because it has a clear model based on
separate layers (the 7-layer OSI model).  Changes on one level (like
10-base-T, to 100baseTX to Gigabit to 802.11a/b/g/n) don't affect the
rest.  Likewise, you can use LDAP, NIS, DNS, /etc/hostnames, or other
means to map IP addresses to hostnames at the application layer-- TCP/IP
doesn't care.

> From my perspective, achieving this objective will require a two
>  pronged approach: 1) Identify the framework dependencies on outside
>  components, and 2) avoid introducing new framework dependencies on
>  outside components. 

This assumes the "framework" is the lowest level.  If the framework
depends on outside components, then the hierarchy has been upset, and
spaghetti dependencies are the inevitable result.  Dependencies HAVE to
be unidirectional, or you never get out of the maze.  IMHO, the biggest
problem with MVC is that it has never seemed to me that the layers are
very well defined.  Everything seems pretty interdependent, and you
quickly get into a rock/paper/scissors kind of analysis, as you
describe.

Is there a comprehensible map of the layers in OFBiz?  All I have seen
is very detailed charts that seem to obfuscate, rather than clarify, the
relationships of the various modules.  But I'm sure I have not seen
everything.  Is there a 30,000-foot overview of the software levels?
 
> The first prong can be accomplished through contributions from people
>  like you - find the dependencies and create patches to fix them.
> 
> The responsibility of the second prong is up to the committers. We need
>  to be more vigilant to guard against introducing new dependencies.

Which requires a clear model of what layer the code under consideration
belongs to, and what are the well-defined layers below it that can be
dependencies.
> 
> Personally I believe it will be possible, BUT it won't be easy. The
>  obstacles to overcome will be getting people to contribute to the
>  effort, and getting committers to avoid introducing new dependencies.

Again, I think we need to reduce the learning curve by providing clear
maps.  You shouldn't need to know everything to be able to contribute
meaningful and error-free code.
> 
> -Adrian
> 
> 
> --- On Fri, 2/5/10, Christopher Snow <sno...@snowconsulting.co.uk> wrote:
> 
> > From: Christopher Snow <sno...@snowconsulting.co.uk>
> > Subject: what a mess! is framework independence ever going to be possible?
> > To: user@ofbiz.apache.org
> > Date: Friday, February 5, 2010, 10:58 PM
> > I'm back to the process of working
> > out how to get a standalone framework running based on
> > trunk, but I have found that the dependencies have got out
> > of hand (if I've understood the code right):
> > 
> > Framework  depends on Themes
> > Themes depends on Content
> > Content depends on Party
> > 
> > The questions I'm starting to ask myself are:
> > 
> > "Is is ever going to be possible to have framework
> > independence in trunk?  Independence in 9.04 is
> > relatively trivial (rewrite security screens) perhaps the
> > most sensible thing would be to do a fork of 9.04 and then
> > back port all framework related commits from trunk? "
> > 
> > Any ideas anyone?
> > 
> > Many thanks,
> > 
> > Chris
> > 
> 
> 
>       


-- 
Matt Warnock <mwarn...@ridgecrestherbals.com>
RidgeCrest Herbals, Inc.

Reply via email to