Do we Need a new Adempiere Architecture

Create your own tidbits here. Drinks on the house. Ash-trays provided.

Do we Need a new Adempiere Architecture

Postby mdsahni » Tue Sep 12, 2006 7:02 pm

Hi Guys,

I just wanna touch upon this topic, as are about to start a new journey, we might as well evaluate our options.

Compiere is maturing [growing] like any our ERP application and I don't wanna see it crack under its own weight.

Many times in past we have seen the core changes and it breaks lots of external works [Jasper is a classic example]

Wouldn't it be good we slice the adempiere in Kernel [Core app] and Add-ons [modules] and the Kernel will not change very frequently [even if does, we ensure backword compatibility] and the majority of community will be developing Add-ons, which could be Process Specific [Advanced Costing, Pricing, etc] or even industry specific [Call Billing, etc]

We can work out some guidelines on how the Add-ons will be developed, so that they can be easily installed or uninstalled on the existing Adempiere Kernel. [XML2AD and 2Pack can form the basis of this]

This is where our AdempiereForge will also play its own role in clustering all the Adempiere Kernel and Add-on projects development at one location adempiere.org.

I'm not a super java architect, but I'm sure we have all among us who can work and make this happen.

Yes this is not easy and we have to see all the aspects but IMHO this will be the single most powerful step we can take to expand the reach of adempiere to unimaginabe levels.

I would really like to hear from all you Gurus out there!!

Regards,
Deep
mdsahni
 
Posts: 21
Joined: Wed Mar 08, 2006 9:35 pm

Postby red1 » Tue Sep 12, 2006 10:38 pm

i agree sir, and if u can help setup those on some server fast, we can start pointing the adempiereforge.org there.

very good points!
red1
Site Admin
 
Posts: 2762
Joined: Tue Jul 06, 2004 3:01 pm
Location: Kuala Lumpur, Malaysia

Postby mdsahni » Tue Sep 12, 2006 10:53 pm

One more benefit I can see now is alignment of Adempiere team!!

Right now we all know Compiere and can [and ready to] work on anything but if have to start working now, the task allocation will be quite messy and without proper division of work and directions, we may end up duplicating and wasting efforts.

If we see in bit deep, we all have strengths in specific areas like manufacturing, accounts, core, etc. and if we divide the app in specific modules, and make One Champion for each module, who'll have more say in taking final decision for that module [though mostly with consensus, but need to have authority, in case].

I'm already working setting up the website [with all other features] and hopefully will give you guys a sneak preview by tomorrow for your comments!!

But most important is to have one real discussion on the Architectural issues and how can we proceed, pit falls and maybe the approach, etc.

regards,
Deep

PS: Even our Forum can be structured based upon Adempiere modules, rather that just one big technical discussion!!
mdsahni
 
Posts: 21
Joined: Wed Mar 08, 2006 9:35 pm

Postby StevenK » Wed Sep 13, 2006 12:27 am

From a developers' point of view, the advantages of having a small kernel and the rest of the functionality as a plug-in are speed in implementation and ease of customization. These are the advantages of modular programming. Of course, a plug-in manager would be essential.
On the other hand, the current architecture is designed a bit monolithic. This has the advantage of ease of development, but if the program gets too big and gets new features, all the unwanted features will slow its execution and the needed effort for the manageability of the code will be next to impossible.
(that's the opinion of my partner, who IS a developer - I am not!)
I think that modularity is a desirable feature, for now, and a **necessary** feature for the future of the project!

Regards,
StevenK
StevenK
 
Posts: 8
Joined: Fri Sep 08, 2006 6:42 pm

Postby red1 » Wed Sep 13, 2006 6:27 am

Can Steven talk more of this model and perhaps draw up some scaff-folding for it? Its attractive idea and makes sense.

Looking forward to it, MDSAHNI! :D
red1
Site Admin
 
Posts: 2762
Joined: Tue Jul 06, 2004 3:01 pm
Location: Kuala Lumpur, Malaysia

Postby Rasmus Luffen » Fri Sep 15, 2006 4:01 am

If u're thinking of new architecture, and bugfix, testability is main focus in t'beginning. work local. break dependencies. JUnit tests.
That is the big challenge.

sincrly
Luffy
Rasmus Luffen
 
Posts: 5
Joined: Fri Sep 08, 2006 5:24 pm
Location: Bullerby

Postby Jens » Tue Sep 19, 2006 7:49 pm

StevenK wrote:...Of course, a plug-in manager would be essential.


Kettle (the Pentaho Data Integration) uses plugins, if you want to have a look how it is managed: kettle.javaforge.com
I'm not shure if it fits to Adempiere (don't know it yet very well).

Keep on going this way.... Jens
Jens
 
Posts: 1
Joined: Tue Sep 19, 2006 6:21 am
Location: Germany


Return to Open Forum

Who is online

Users browsing this forum: No registered users and 2 guests

cron