Re: [Digir-dev] Question on DiGIR Darwin Core and uniqueness of Catalog Numbers

[Robert A. M. <ra...@cs...>]

Forgive me if this point has been covered in previous discussion I may
have missed.

Hannu Saarenmaa wrote:

> 
>>>
> [...]
> 
> Don't worry, replacing IH is not in the cards.  Rather, we are talking 
> about additional infrastructure that could be used to do the following 
> programmatically (via web services):
> 
> [...]
> 
> 
> - Find out duplicate records and various versions of the records. 
> - Link physical specimens to the electronic records.

Do these two cover the scenario in which a specimen moves, either by
acquisition or possibly by loan, and the new holder assigns a new LSID?
At least in the acquisition case, the new holder needs(?) to do this
since it is in a new collection and institution, but how is this new
LSID associated with the old LSID, which may appear as a reference in
data held somewhere in the world, possibly even in the collection
records of either the old or new holders of the specimen?

By  taxonomic analogy, I suppose this is the question: "How is LSID
synonomy memorialized"? It also has the same problems as does taxonomic
synonomy, namely the date of synonomization must also be knowable if
references are to be correctly resolved.

Or is the acquiring institution required to keep the original LSID on
the specimen and deal internally with the consequences of having LSIDs
that do not reflect its own identification of the specimen? In such a
case, all the syntactic structure of the LSID will not have much value
except for historical interest---and not even much of that if the
specimen moves a second time. Put another way, in this case, the LSIDs
must be regarded as absolutely arbitrary, must have no interpretation
whatsover attributed to them, and the scheme is only a way to manage
avoiding central administration of issuance of LSIDs. Not that this
would be such a bad thing...


Bob

p.s [for wannabe networking geeks. Among the readership, this might be
just me, Hannu, and Donald...] FWIW, the two alternatives---permitting
synonomization or forever frozen  ID is, in current internet
architectures, the difference between IP addresses and MAC (roughly
"ethernet") addresses. The latter come from blocks of numbers assigned
to manufacturers of network hardware. The former are assigned by
insitutions using a scheme like hierarchical URIs. If you move your
network interface card to a different computer, its id is not supposed
to change, whether or not the IP address of the old computer is assigned
to the new. [DNS addresses are a whole different matter even though
their construction superficially appears similar to LSIDs. They are
always resolved at the client side, albeit with the implicit help of
remote services. That's because all IP connections are by IP addresses,
not DNS addresses---which as TDWG users sometimes notice need not even
exist]. This model requires /providers/ to internally and transparently
resolve the (mutable) IP address to the (immutable) MAC address. In
fact, both on technical and security policy grounds it is generally
impossible, pointless, or a really bad idea for the MAC address of a
machine to be discoverable by computers not on the same wire.

It is conceivable there is merit in examining the OSI 7-layer network
model as an analogy for insight into the services Hannu laid out, and
possibly others. The question at hand would be expressed as: do LSIDs
serve the function that IP addresses serve, or those that MAC addresses
do. Certainly LSIDs have to be knowable beyond the provider, so

-- 
Robert A. Morris
Professor of Computer Science
UMASS-Boston
http://www.cs.umb.edu/~ram
phone (+1)617 287 6466