What's the difference between the different Jersey test containers?

Question

According to http://jersey.java.net/nonav/apidocs/latest/jersey/jersey-test-framework/jersey-test-framework-core/com/sun/jersey/test/framework/JerseyTest.html users may choose between two kinds of test containers:

1. Low-level container that does not support servlets
2. Web-based container that does support servlets

I have the following questions:

1. What's the advantage/disadvantage of each kind of container? For example, I assume low-level containers are faster but are missing some sort of functionality.
2. What is meant by the fact that low-level containers do not support servlets? I thought that JAX-RS is based on top of Servlets. What aren't low-level containers able to do?

UPDATE :

Upon further research I discovered that low-level containers are faster but they really run without servlets. When clients make requests,a the appropriate Resources are constructed and the response is passed back to the client, all in-memory. This is similar to running an embedded database versus a conventional network-based JDBC connection.

There is no mechanism to supply a ServletContextListener, register servlets or filters. I'm guessing the request scope isn't supported either.

My question is, can you really test anything meaningful under these restrictions?

Answer 1

"... can you really test anything meaningful under these restrictions..."

You can test your business logic, and do it quicker.

(Would like to see this fixed though: JERSEY-622 . The in-memory test container doesn't support Jackson, ie the POJO-mapping feature of Jersey.)