1) problems in 3tier scenarios: i have to attach the entity on the server, load the data again from db to get the changes or save all properties (yuck).
2) misleading API. I do not like apis that do not do what their name says:

GetById() = look in the cache first and than in the db (yuck) depending on the configuration.
I would preffer: GetById(x, Options.UseCache); that overrides any configuration

It does not know how to map interfaces not classes

and so on…

it is a very good ORM if you want to have a surprise every day.

NHibernate forces you to think of the problem in terms of objects which is a great departure from what you normally see in the field where domains are perceived in a procedural style.

You do need an NHibernate champion on the team.

Other practical considerations are:
1. Effort of mapping the domain. It’s ok to look at an example, but then most application scenarios go beyond the example. There is the Fluent approach to mapping, but this appears to a non-core feature. Then there is NHibernate Mapping Attributes which has it’s own issues.
2.The documentation available seems to have a Hibernate-esque feel to it; if you are from a Java background then no probs. Otherwise this can be a hindrance.
3. The stack traces tend to be somewhat non-informative, which can be painful when your app has blown up.
4. Tool support- it’s great to have an NHibernate Profiler but I would expect to at least have a free version of the tool. I now have to resort to using the SQL Profiler to view what’s going on in the background. I can use the Show SQL option. I guess making tools costly raises the barriers to entry in the NHibernate game.

I definitely learned that it’s a feature-packed ORM, but the first wall of getting what it is and how you can start is hindering me that doesn’t much time and motivation to try it.

I currently use ECO from CapableObjects which does all that I want, and has getting started docs too (I wrote them), but it’s always good to learn as much as you can from as many sources as possible so I have recently decided to “bite the bullet” and buy NHibernate in Action. Once I have read that book I will finally make an informed decision rather than the uninformed one I have made to date.

The lack of docs though is just an absolute killer for me. I don’t have enough free time to learn something before discovering if it is any good or not, I need to skim over docs and judge before I dedicate the time.

1. Lack of good documentation. This means both in-depth documentation of everything, but mostly getting started docs. Things like step by step setup, examples, and best practices all located in one place.

2. Lack of up to date, easy to use code gen templates (T4 / CodeSmith / MyGeneration / etc)

- Won’t try it because i use something else that i’m perfectly happy with
- Won’t try it because i’ve heard too many stories of projects having problems with it

I do watch to see if there is something better. I’m always evaluating. Your PR and my experience is that NHibernate is hard to set up; when your project has to have lot’s of tutorials none of which actually satisfy the problems of a newbie learning to use your project…
Also, I love the simplicity of setting up SubSonic (my personal choice).

But its very hard to see how to migrate from this to NHibernate. (esp. because have no experience with NHibernate).

A big help would be a good technology independent data access architecture, we could migrate to first. Then replace our current stuff with NHibernate second.

In the project I’m working on we use immutable DTO objects to move data from the DB to the domain. All the properties and private fields are read only so they can only be setup through their constructor. I couldn’t seem to find a way to get this to work with NH. As far as I could tell from the docs and my own spike I had to add a default constructor and make the fields writeable – not really a probem, but hardly persistance ignorance.

Having said that we’re about to start a new project, and I’m very keen to use an ORM right from the start

Plus the unfortunate attitude we live in still that everything data related should be extracted using stored procedures and datasets.

yeah the documentation is indeed a problem right now… while most of the documentation for 1.2 is still accurate, the 2.0 specifics haven’t been documented properly yet

@Joachim

personally, i don’t like visual designers at all but i can see why the lack of it would put some people off, especially beginners.

having said that, i do believe that the manual mapping approach is much more productive in the long term than any visual designer could be