Well, if you look at it the right way, using a sparse file as an iSCSI extent might be considered "thin provisioning." Consider this:
total 119085856
drwxr-xr-x 2 root wheel 3 Sep 24 08:38 ./
drwxrwxrwx 18 root wheel 37 Nov 4 01:08 ../
-rw-r--r-- 1 root wheel 1099511627776 Nov 8 00:42 extent0
The total space used within the directory is nowhere near the claimed filesize; this is because the file was created using truncate and is being slowly allocated as it's being written to. The same thing works with CIFS of course. This is of only limited usefulness, but you can indeed allocate a file that's nearly infinitely large without regard to your actual available storage.
Now, for something that is more practical, the use of multiple datasets in ZFS can give the appearance of thin-provisioning-like behaviours. If you have a 1TB pool, and you create "dataset1" ... "dataset4", each dataset by default reports the remaining pool space as its own free space, but in reality the free space for 2, 3, and 4 goes down when stuff is written to dataset1 (etc).
It should be obvious that in all cases, you cannot actually STORE more data than you have space for. Deliberately setting up a scenario where the reported available space is greater than what you have requires some thoughtful design decisions; for example, it'd be exceedingly bad to rely on thin provisioning at the storage level in order to store ESXi VM's that in turn were thin provisioned; you have an exceptional opportunity for extreme badness to happen if the disk fills. ESXi pretty much freaks when it suddenly can't write to something it ought to be able to write to...