[eml-dev] customUnit database...

[Matt Jones]

Hi Mark,

This sort of question should really go to eml-dev as much of this work
was done by others, not by me.  Nevertheless, I'll chime in as best I
can and hope Chris, Chad, and others that helped develop this comment as
well.

First, we (eml-dev) have discussed and would like a community-based
mechanism to separate unit definitions from the EML spec altogether.
This is because we have found many problems and omissions from the
existing set (see bugs 1000, 1605, 1634, URLs listed below), and think
it would be best to separate the unit defs from the EML specification so
that we don't have to revise EML just to add new unit definitions.  Such
a move would require some review of unit submissions -- one thing we can
not do is eliminate or redefine old unit definitions that might be
referenced in various metadata documents -- they will need to have a new
identifier.

In addition, for this to work for EML generally the system can't be
restricted to a single community like LTER, but what you are proposing
sounds like just the sort of thing the broader eml-dev community has
asked for.  We have partially descibed our needs for a repository in
bugs 1000, 1605, 1634, and for general enumerations in bug 1197 in
Bugzilla. Would you be willing to generalize the task for the benefit of
the whole community?  I think the critical factors would be:
institutional independence, persistance of unit identifiers, a web
accessible UI for searching, a machine-accessible API for searching and
getting the list of units and their definitions, and possibly a review
process.

Regarding your specific questions:
The base dimensions come directly from the NIST web site
(http://physics.nist.gov/cuu/Units/), which was our guide through most
of this.  We added dimensionless because there are some tricky issues
with some basic measurements that aren't really accomodated well in the
NIST view of the world -- in particular counts and proportions and
ratios, as well as quantities like pH.  This stuff, especially counts,
starts to get into semantics and philosophy of measurement pretty
quickly, which made it hard to pin down in the EML release.  We've been
thinking about it a lot more in SEEK and still have fundamental problems
modeling these concepts.  So, I would treat any 'dimensionless'
quantitiy in EML as essentially experimental but necessary to make
progress for the time being.  I'm not really sure where it goes.  We
made it a dimension so that it can be referred to, and we made a
unitType and unit for it, but didn't define it as an SI unit as SI
doesn't really recognize such a thing (its fundamentally different than
a meter).  More discussion is needed on this.

Regarding charge: it appears to me that this is a bug in EML.  Somehow
the fundamental dimension "current" was put in the dictionary as
"charge", and this is a problem.  The seven Base SI units and their
fundamental dimensions are listed in NIST's table 1
(http://physics.nist.gov/cuu/Units/units.html), and should include
length, mass, time, current, temperature, amount, and luminosity.
Charge should be a derived unit as you describe (sA). We need to fix
this asap.  I've entered a new bug 2083 for it.

Regarding angles: We identified the problems a long time ago but nobody
has fixed them.  The issues are described in bug 1605.  We need to fix
this asap.

If you read the bugs that I have listed you'll find many more issues
than just the ones you have raised.  The really thorny ones get into the
semantics of measurement (e.g., what a 'count' is and how it relates to
the SI 'mole' and dimension 'amount').  Thanks for the discussion prompter!


Bug URLs:
http://bugzilla.ecoinformatics.org/show_bug.cgi?id=1000
http://bugzilla.ecoinformatics.org/show_bug.cgi?id=1605
http://bugzilla.ecoinformatics.org/show_bug.cgi?id=1634
http://bugzilla.ecoinformatics.org/show_bug.cgi?id=1197
http://bugzilla.ecoinformatics.org/show_bug.cgi?id=2083

Matt

Mark Servilla wrote:
> Hi Matt,
> 
> I hope all is well.
> 
> I have been working with Karen Baker and others on the development of a
> custom unit database (to search, add, share, and manage new units in
> addition to those in the eml-unitDictionary.xml).  In doing some
> research on the origin of the unit dictionary, I notice that the
> fundamental unit types do not correspond 1:1 to those defined as base SI
> units (but, they do match with those defined in the STMML lineage).
> Specifically, fundamental types include dimensionless (probably
> necessary), charge, and angle; where base SI does not include
> dimensionless, define charge as derived unit (sA), and clearly
> distinguish angle as either plane (radian) or solid (steradian), also
> derived.  In addition, the fundamental types do not include current,
> which is a base SI unit.  Should there exist a closer relationship
> between the two groups, or am I missing something?  I apologize if this
> is a revisit to old news or if I am really off the mark.  Thanks!
> 
> Sincerely,
> Mark