Hi Magnus
Am not completely clear what your question is, but..
Making ldap (and database) queries from a website is stateless. Each call will require authentication - as in the rails generated database.yml file.
You could have a model that authenticates the users against ldap and caches relevant information - role, name, etc, etc. So it's available for the lifetime of their session.
At a large bank, I worked on an active directory reporting system. There was *some* real time lookups, but most of the reporting work was staged a sql database. We had continuous jobs loading the database from the active directory.
The data wasn't up to the minute fresh, but it was good enough for our needs. The advantage was that using a sql database was a lot easier to create the complex reports that were needed.
For you, it will depend on what data you need - only ldap objects? other data created by the users?
If the website is driven by a database (as opposed to ldap), you'll also have a much smoother development with ActiveRecord, etc
hth