sftp doesn't count. (Or does it ?)
I found a Savannah ticket that tells me I can use lcg commands by prefixing them with 'lcg', e.g. 'lcg lcg-ls' or getting a glite environment by doing 'lcg bash' and the lcg-ls.
Is this documented anywhere ?

(a) Copy my credentials onto the cloud node (here the DOC one).
mkdir .globus
copy usercert.epm and userkep.pem
(b) First try:
[dbauer@host-10-40-1-66 ~]$ lcg bash
[dbauer@host-10-40-1-66 ~]$ voms-proxy-init --debug --voms cms
[snip]
Your proxy is valid until Thu May 26 04:55:29 2011
[dbauer@host-10-40-1-66 ~]$ lcg-ls --vo=cms --connect-timeout=300 srm://gfe02.grid.hep.ph.ic.ac.uk/pnfs/hep.ph.ic.ac.uk/data/cms/store/user/dbauer
/pnfs/hep.ph.ic.ac.uk/data/cms/store/user/dbauer/tutorial2011
Good.
[dbauer@host-10-40-1-66 ~]$ lcg-cp --vo=cms --connect-timeout=300 srm://gfe02.grid.hep.ph.ic.ac.uk/pnfs/hep.ph.ic.ac.uk/data/cms/store/user/dbauer/tutorial2011/exercise_33_2_qAr.root exercise_33_2_qAr.root
[dbauer@host-10-40-1-66 ~]$ ls -l
total 12
drwxr-xr-x 15 dbauer dbauer 4096 May 12 13:38 CMSSW_4_1_5
-rw-rw-r-- 1 dbauer dbauer 4594 May 25 17:09 exercise_33_2_qAr.root
Even better.

Can I shove data back onto the SE ("Storage Element" in grid speak) ?
lcg-cp --vo=cms --connect-timeout=300 --sendreceive-timeout=3600 exercise_33_2_qAr.root srm://gfe02.grid.hep.ph.ic.ac.uk/pnfs/hep.ph.ic.ac.uk/data/cms/store/user/dbauer/exercise_33_2_qAr_copy.root

That should do it.

---