Hi All, I just wanted to confirm what Steve said: We are talking about the CESM public data server (not the CESM development server), which ACME is going to continue to use - all agreements are in place and instructions are on the ACME quick start guide. In fact, the CESM data server specifically asks users not to clone the entire collection - only download what is needed, which is why the scrips work the way they do. The ACME scripts will first check the ACME data server, so all ACME specific data will go there. From my point of view, this is a resolved issue and documented in confluence. The only issue we need to resolve now is NERSC specific configuration detail: should ACME have it's own inputdata directory, symlink into the ccsm1 inputdata directory to duplication, or work out an agreement to use the ccsm1, or come up with some other arrangement. Mark -- Mark Taylor, [email protected], +1-505-284-1874 1446 Multiphysics Applications Sandia National Labs, Albuquerque NM 87185-1320 From: Steve Goldhaber <[email protected]<mailto:[email protected]>> Date: Monday, June 8, 2015 at 12:00 PM To: Hans Johansen <[email protected]<mailto:[email protected]>> Cc: "Taylor, Mark A" <[email protected]<mailto:[email protected]>>, "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: [EXTERNAL] Re: [ACME-SE] Request to download missing inputdata on Edison I'm not sure it is generous (except for the hosting), it is just part of the normal open science process (especially with the new directives on open publication<https://www.whitehouse.gov/blog/2013/02/22/expanding-public-access-results-federally-funded-research>). I'm sure ACME will release the relevant input data sets when it releases a new model version. These models are pretty useless without the input data. --Steve ----------------------------------------- On Mon, Jun 8, 2015 at 11:50 AM, Hans Johansen <[email protected]<mailto:[email protected]>> wrote: Great! I didn't know that, how generous. I'm sure that's not bi-directional, though - so ACME inputs would have to live somewhere else with acme access controls. Hans On Mon, Jun 8, 2015 at 10:40 AM, Steve Goldhaber <[email protected]<mailto:[email protected]>> wrote: On Mon, Jun 8, 2015 at 8:50 AM, Hans Johansen <[email protected]<mailto:[email protected]>> wrote: Yes, that sounds right (guessing how linkdir_tree works). However, Rob's right - my point was to get away from the CESM-based "login in to ORNL / NCAR to get required input files if you want to run this case" process. I found it very frustrating on Edison, and it clearly is bending our collaboration rules. It's hard to automate and doesn't speak to repeatable ACME runs. One issue I don't understand is your comment about collaboration rules. All of the CESM input data (the https://svn-ccsm-inputdata.cgd.ucar.edu/trunk/inputdata/ repo) are released to the public and are not part of the ACME/NCAR agreement. --Steve