[School-of-data] Is data open if you can not create derivatives?

[Tom Roche]

Michael Bauer Sat, 12 Jul 2014 10:11:35 +0200
> Good metadata is something that's incredibly valuable and not commonly found.

Too true.

> If any of you want to write a rant about that - and what kind of [metadata you want]

Discussion of infrastructure (APIs, containers, file formats, repositories, etc) that encourages metadata provision might also be valuable. E.g., IMHO, netCDF makes annotating data relatively easy and efficient, and its API/framework makes get/set-ing metadata straightforward. That doesn't prevent folks from putting crappy netCDF "out there" :-( but it does provide some minima (e.g., one always knows what the dimensionalities are, even if dimension semantics are not always clear), and it facilitates shaming of defectors (because if one *doesn't* annotate, there's really no one else to blame). Some netCDF tools (e.g., NCO) are also quite good at recording data operations to metadata (using global attribute=HISTORY) which IMHO deserves wider adoption.

FWIW, Tom Roche <Tom_Roche at pobox.com>