I admit that partitions must be contiguous. But the only reason one normally only changes the end address of a partition is because the contents (file system) can only be adapted at the end (when at all). So when that (content lost) is not a concern you can change what you want.
When you have e.g. sdn1 of 4 GiB and imediatly after it sdn3 of 2 GiB and one wants to make sdn1 2 GiB and sd2 of 3GiB, using a tool like fdisk, you will probably remove sdn1, remove sdn2, create a new sdn1 and a new sdn2 and then write this new partition table. You could of course then say that you have shrink sdn1 at the end and expanded sdn2 at the begin. BUT IMHO that is not the way to talk about this. These tools always write a complete new partition table. Even when names (sdn1 and sdn2 in this case) are the same as those used earlier, all is fresh.
And it could be that on sdn1 there is still a file system that only fills 2 GiB. And that will be still usable.
