Most of the customization energy is built into the tools now. Compiere calls its customization process "model driven," which means that you just start adding the columns to the tables in the data models and Compiere does most of the rest. Adding a field to a form means filling out several additional forms, an act that left me wondering whether forms begetting more forms meant that they were a life force all of their own.
One piece of Compiere documentation promised that the customization process required no "error-prone procedural programming," which I thought was a pretty accurate description. Adding new lines to the forms and building new rules for them is programming, it's just not at the grungy Java level. The developers took their ERP mechanism and applied it to managing the source code itself.
[ Need a Linux that can squeeze into a pen drive, resurrect old hardware, or create a security or storage appliance? See "Specialty Linuxes to the rescue" and "More specialty Linuxes to the rescue." ]
Turns in the open road
Which should you choose? The most important criteria is one that can't be answered in a general review like this one. Each business has different needs, and it's hard to assess whether any of these even match the general workflow. The best tool with the most features can be a horrible fit for an organization if its internal architecture depends upon some structure to the workflow that doesn't work with your business. Sometimes it may be more expensive to shoehorn a large operation into a standard product like these than it would be to write custom software from scratch.
While it's doubly foolish to mix stereotypes of people with generalizations about software, I think that Openbravo is like guys in the warehouse: They want to fill orders and deliver the goods. SugarCRM is like the hail-fellow-well-met hand shakers who know that much of the sales process depends upon getting in front of people and keeping track of their needs. Compiere is a bit of a mixture of both, but is dominated by the spirit of moving material through the warehouse.
But just as humans can move between roles, software can too. These systems are all open collections of database tables and rules for editing them. If something is done with one of them, it can be rewritten for any of the others. There's no reason why Openbravo can't act more like the smiling, backslapping SugarCRM, and there's no reason why SugarCRM can't add the right plug-ins and start working in the warehouse.
In all three cases, customization is simple enough that it is becoming harder and harder to choose to build your own. It would be pretty silly to write your own back office in Java or C++ today, not only because so much functionality is already there for the taking, but because these systems are close to being languages themselves. The combination of the open source development and the rise of a plug-in architecture means that it's easier and easier to modify what exists. These apps were designed to accept the contributions of everyone, and this flexibility is baked into the entire development cycle.
Still, these three projects are a far cry from the image of a programmer-led commune where everything is shared equally. These are professional companies first and foremost, and they sell a product that they also happen to give away. Will the free, open source edition meet your needs? It may not be simple, but a skillful programmer should be able to install all of them and get them storing data in a reasonable amount of time.