Uncertainty over UDUNITS support

[paulmillar]

The NeXus type definitions include anyUnitsAttr, which is used as the set of valid values for a units attribute. This attribute is used explicitly in the NXdata and NXparameter base classes, in the NXcanSAS application profile, and in various contributed definitions. The anyUnitsAttr definition is based on UDUNITS.

A recent announcement has announced that NSF will have to "pause [...] most operations" due to staff undergoing a furlough. The announcement also mentions that the NSF's "documentation, downloads, and software artifacts servers will remain online to the extent possible."

This announcement creates uncertainty in how UDUNITS will continue. As a reaction to this uncertainly, a GitHub fork of UDUNITS has been made, to provide guarantees that information is not lost.

This issue exists to track the UDUNITS user community's reaction to this uncertainty and to allow the NIAC to update NeXus, as appropriate.

[prjemian]

There exist now some alternatives to UDUNITS. NeXus could revise its basis to a standard or software package that provides unit conversions. One example is pint. There are others.

[ggoneiESS]

Fantastically timed, but this morning there has been a change back to normal - https://www.unidata.ucar.edu/blogs/news/entry/nsf-unidata-return-to-operations

The SCIPP group here at ESS use https://github.com/LLNL/units/ to validate units - which so far has been fully compatible with UDUNITS that the raw data uses units verified by that package.

I think Pete's idea of making the standard broader somehow is a good one, but I'm not sure mandating any particular package is the best way.