Stephane
The scan is scanning the whole index (the whole table in the case of a clustered index) whereas a seek goes straight to where it needs to go to find the data. Therefore a seek is usually desirable. I think the reason for this is that the query optimizer can't use the clustered index to work out the "<= ipTo" part of the "BETWEEN ipFrom and ipTo" clause. Therefore I would recommend you create a non-clustered index on ipTo (assuming that your clustered index is created on ipFrom and ipTo in that order).
John