R – How to move documents and list items in Windows SharePoint Services

Object model is better as you can gain access to additional features and the full detail of the list items, such as the version history.

The object model is also better for performance (as long as you dispose() your spsite and spweb objects properly).

The Sharepoint object model has some differences between 2 and 3, but if you look at the reference for v2 then it will also work fully with v3.

The web services have not changed at all between v2 and v3, which explains why they do not have any new features of v3.

The reason the object model will win on performance is that you will not be serialising the data as Xml and then transmitting a large chunk of Xml, and then deserialising the Xml. The object model spares your memory and bandwidth.

What you need to do is not supported by SharePoint, along with a many other cross-site and cross-farm operations. Let us hope this will improve with future releases.

You should check out products from AvePoint. They have many of these capabilities and more, but are reputed to be expensive. They provide an API if you do not have qualms about 3rd party dependencies. You mentioned that DocAve is single-site only but that is not the case. You might want to contact the company to learn more.

The programming approach requires you to build all your own scaffolding using the object model API or the web service API. The web service API is easier to use and allows your code to be more location independent, but it lacks too many features if you need a fine level of control. In addition to files and folders you need to worry about users, content types and any other list item your application needs.

The object model approach requires the code to run on a machine that is a member of the farm. This means you will need a local "agent" in a multi-farm deployment and create the associated distribution logic.

All this may sound daunting and while it is not trivial but can be done.