[Taxacom] Showing the importance of natural history collections and the networks they are building

[Richard Pyle]

Hi Quentin,

> do the DOIs (or other identifiers) allocated to specimens refer to the
physical
> object or the database entry describing them?

This is exactly the issue I was trying to get at.  I think it depends on who
is in charge of issuing the identifiers.  If it's a collections person who
is used to dealing with catalog numbers, and who is thrust into the role of
informatician (e.g., due to lack of funds to hire an informatician), then
the identifiers are intended to represent the physical (really conceptual --
but that's another discussion) specimens.  However, if the person in charge
of issuing identifiers is from the informatics world (without a background
in natural history collections), then often the intention is to assign the
identifier to the data record.

The real problem is that this fundamental dichotomy exists, but only a very
few people on the front lines even realize the dichotomy exists.  And even
fewer people take the time to make it clear to the outside world what their
identifiers actually represent.  It's fashionable to bash LSIDs these days
(perhaps justifiably), but one thing the LSID exercise forced our community
to confront (besides the world of RDF) is this distinction between digital
objects and conceptual/abstract/physical (i.e., non-digital) objects;
particularly in terms of the distinction between "data" and "metadata"
(sensu LSID spec).  Other than media files (images, videos, audio files,
PDFs, etc. -- and *possibly* gene sequences...), almost all of the stuff we
care about in biodiversity informatics is in the realm of
conceptual/abstract/physical.  Thus, the trend has been towards minting
identifiers for such.

I'm not saying that identifiers for data records isn't important (it *is*
important -- especially in the iDigBio space).  I'm just saying that our
community hasn't been particularly diligent about being up-front and clear
about what our identifiers are intended to represent.

> I would have thought it was the former, in which case versions are rather
> irrelevent. I tend to agree with Rod, that versioning is not that
important.

No, it's not so much about record versions, as it is about *dataset*
versions.  It's less about what metadata was associated with each record
than it is about what records were there.

Aloha,
Rich