Note the difference between open access and open data. Peter's point generally is that data should be open data, not open access. NIST ms spectra are open access, but not open data.

While I understand that data curation is not cheap, current chemoinformatics and IT technologies are of that quality that the curation can be done at the chemist's bench, which has to do it anyway. A lot of money is currently waisted on people/scientists retyping, redrawing what another might have meant. The current pipeline is inefficient, and doing it right makes open data economical feasible. Peter's efforts in changing the crystallography data in the UK shows that this is quite realistic.

A CIF-like format for organic chemistry synthesis will do the same for other fields of chemistry too.