The single query Eclipse pagination model is claimed to be more efficient compared to a two query method, but I'm not so sure.This class might seem inefficient, exactly because it doesn't use mentioned database-specific SQL commands. However, for one very good reason, that isn't really true: it is almost always necessary to know the total number of pages in a query result - for example to be able to show a page navigator - and as this value is computed from the total number of rows in the query, the only way to get this number is by executing the full query. This completely eliminates the possible gain in efficiency when using LIMIT- or SELECT TOP-clauses, because in that case an additional (counting) query must be executed. Also, consider that most queries specify an ORDER BY clause, and remember that a DBMS can only compute this ordering by examining all rows in the result, even when just the first 10 are selected.
Say, you have a mature discussion forum with hundreds or even thousands of topics and many cols to get in the query, that's a huge result resource to create.
With the two-query LIMIT method, the first (total rows) query creates a much smaller result resource (a single column, probably an integer ID) which can be destroyed once you've got the total number of rows. The second (LIMIT) query, with just the rows for a single page, also creates a much smaller resource.
I don't know if there is any big advantage either way in terms of speed, but in terms of memory use, the two-query method would seem to have the upper hand. The advantage would depend on how many columns you need to fetch, and how many rows are in the table.
Eclipse notes state that avoiding use of LIMIT allows the class to support more databases so I guess the two query method isn't as portable.