For this, it’s kinda weird… why would you build the ability to use several different OR/Ms? Your abstracting on abstractions already… well, I guess that isn’t necessarily a bad thing – but I guess the question would be why? You are removing the point of what OR/Ms are anyway. A quick way to generate concrete classes from the relational database world. Then you’re going to add another container class that talks to that concrete class?

Doesn’t make sense to me. Overly complicated and unnecessary. If you want to really do what you are talking about then you may want to take a look at db4o.com

Also, if you like using “dumb” containers, subsonicproject.com does make use of “Repository” methods as opposed to ActiveRecord. Very versatile product. There is a reason M$ is funding the project (although maybe not directly…)

The first priority is to get the data manipulation out of the presentation tier. On top of this, it is also important to have a loosely coupled layer of abstraction.

So what ORM’s support this easily and also have great features?

My opinion. These great ORM frameworks are simply awesome when your building quick projects that no one other than yourself will extend, but when you get to the point where millions of others will be using your could, you want to keep them in mind.

I think we should place all the data manipulation in one place, maybe an implementation of an interface. That way, when I call the method, I don’t care what ORM or method is used to get the data. Now, if users want to switch ORM’s, they that ability without breaking the application (unit tests would be nice though).

@MrTea – I understand. It just rubbed me the wrong way (his post).

It was written by the Mindscape guys, and it has an interesting data access pattern using IUnitofWork and Repository patterns. It may be a related concept to lightspeed.

Well I think DI/IoC is a pretty mature pattern/practice. I don’t think anyone can claim intellectual copyright on it

But the point stands, I have used Subsonic in a losely coupled way in several projects and it hasn’t missed a beat for me.

@MrTea – I am glad you found that article useful as Rob pulled the key to that solution (DI/IoC) off the dashCommerce forums where I posted it.

@Mike – I don’t disagree with anything that you said – all I would add is that feature requests cover a wide spectrum and I think if you look at the LightSpeed framework – I don’t think you’ll find it sloppy or catering to the whims of forum members. This could not be achieved with all feature requests, but for ones that make sense, I think the value add is significant and does wonders for customer satisfaction.

@cowgaR – I will probably be taking a fresh look at NHibernate as well. I looked at it back in 2005 and it was a little rough around the edges to me, but I’ll have to re-evaluate it and see how the experience is. Whatever I find, if I think it will be useful, I’ll post it up here.

Is this process sustainable as they gain more and more users and more requests? No. And are they thinking about the long term goals of the software or are they appeasing a current user?

I run a project, Architecture Rules. I have only a few users, so I want to keep them happy. I recently had one user ask to add some code that would really have been the opposite of what I was looking for the project to do. If I appeased him, the project would be confusing to new users, would make the code more confusing, and overall just not be good for the project – but it would have landed a new active user in my small community. It’s not worth it though. I hope these guys didn’t trade one new user today for hundreds of users down the road.

Hey. I’m going to have to turn this into a blog post or something.