Performance is one thing I can't compare. Ease of use is kind of hard to get an idea about without at least some experience of them both. I've got a feeling DO is going to win here, but I've got nothing to compare with. (Nightmare would be that my database and my code goes out of sync and I have to manually figure out what's not working. A failed upgrade on many, many sites isn't popular..)

Code generators, I see the advantages of not having extra tools, source files etc. to take care of, but for me, it's not that important. It can be justified?

Ara, thanks for your answer!

Yes, ease of use is, likely, the most important factor. DO4 has some pros and cons here:

• Its initial usage is pretty easy;
• On the other hand, the deeper you dive into it, the more new aspects you open. In many cases we offer different approaches to common problems, but each of such decision is justified - i.e. we were carefully choosing the approach in this case. So you should take into account that you'll finally learn a lot during its usage.

DO differs from the most ORM tools mainly because we are adepts of extremely rich domain models. Compared to other tools on the market, DO is like WPF vs other GUI frameworks: it allows to build very complex business logic with ease without caring much about such stuff as saving the changes, lazy loading, rollback handling and so on; moreover, there is very rich entity life cycle events (see e.g. Session.Events).

Also, DO is built for modular development: you can write unified services (that may rely on their own persistent entities) providing global functionality for the whole Domain.

Many of our own core services are based on this approach - try to reflect VersionValidator or OperationCapturer. Even such a large piece as DisconnectedState is very loosly coupled with everything else - in fact, it is almost 100% extension (it uses some internal stuff mainly because we could afford this; + we added such properties as Session.IsDisconnected to make it more usable).

So if you like the idea of rich domain models, DO is for you. On the other hand, if you're adept of anemic domain models, you should take a look @ NH or EF.

And about new LLBLGen Pro:

As far as I can judge, currently its only benefit is mapping designer. The ORM framework there is ~ 5 years old - i.e. it's simply far from being contemporary.

So if I'd consider using LLBLGen Pro, I'd almost 100% avoid the ORM it offers. I like its designer, but completely dislike the way its ORM framework is designed. Just compare their class/namespace reference to what's shipped with DO (EF, NH). I can't imagine why Frans put nearly everything into a single namespace.

Also, I hate dealing with tons of generated code violating DRY everywhere. Even generated code must be beautiful & simple to read; if template author simply don't know how to extract the logic to methods of base classes, that's sad.

Why I write about this: ~ 1 year ago we tested its contemporary version @ ORMBattle, and I what I remember the best is that generated code there was really huge - about 1 Mb of C# code for our Northwind model. It was nearly 10 times more than any other tool had produced, and that was really surprising.

So if you consider LLBLGen Pro, personally my opinion is:

• Its designer looks really nice and convenient; likely, it's way much better then what you'll have with EF by default. Also note that using EF designer with 200+ types currently is almost impossible - designer may "hang" for several minutes handling such simple action as scrolling. Most likely, LLBLGen Pro handles this better.
• Everything else has not competitive advantages (try to name any :) ).

And finally, I'm writing a large article comparing DO and NH - if you'll wait for few days, it may help you to make a decision.

I think I'm convinced, but that article sounds interesting anyway :)

The article publication is delayed, but I can share an early link for feature-based ORM comparison we're working on:

The most comprehensive feature-based ORM comparison ever :)

The document is a bit incomplete yet; there can be some mistakes (it wasn't checked by community yet). But it is quite comprehensive (~ 270 features).

A copy of this document is edited by ORMBattle.net participants, so shortly there should be its version including all the tools we tested.

Don't forget to check "Remarks" & "Sources" sheets.

Btw, this explains why there is still no article - I want to publish a really good one.