Hi Tim,
What you outline below (a service that issues community-wide actionable and persistent identifiers) is what I have been advocating since before the first TDWG/GBIF GUID workshops. I still believe it would be a very useful service; however, it seems we are really talking about two different things here.
The first, which is what you outline, is a service that mints new identifiers for objects that do not already have good identifiers.
The other service, which is what I was addressing earlier, is a centralized mechanism for cross-linking *existing* identifiers to each other. Several organizations already have an internal system for doing this. I already mentioned the GNUB version (which allows the little icons of related records to show up on ZooBank pages). EoL also has one(e.g.: http://eol.org/pages/992573/resources/partner_links). So does NCBI (http://www.ncbi.nlm.nih.gov/projects/linkout/). What I think we REALLY need is a single, centralized system that manages cross-links among identifiers and (separately) identifier dereferencing services. The reality is that we already have MANY identifiers minted for the same object (e.g.: http://zoobank.org/2C6327E1-5560-4DB4-B9CA-76A0FA03D975) – and, sadly, there will no-doubt be more redundant identifiers minted in the future.
While I think it would be GREAT if GBIF did offer a service to mint proper/actionable identifiers for the community, ultimately this may end up representing “yet another identifier”. It kind of reminds me of that standards joke: http://www.howtogeek.com/geekers/up/sshot50509a6b8cb11.jpg Just replace the word “standards” with “identifiers”, and we’re in the same place.
And I’ll make this point one more time: Almost all of our problems with identifiers revolve around the fact that we have drank the TBL Cool-Aid and assumed/insisted that we conflate the role of object identification with the role of metadata dereferencing (i.e., “actionability”). Sigh…..
Aloha,
Rich
Richard L. Pyle, PhD
Database Coordinator for Natural Sciences
Associate Zoologist in Ichthyology
Dive Safety Officer
Department of Natural Sciences, Bishop Museum
1525 Bernice St., Honolulu, HI 96817
Ph: (808)848-4115, Fax: (808)847-8252
email: deepreef@bishopmuseum.org
http://hbs.bishopmuseum.org/staff/pylerichard.html
From: tdwg-content-bounces@lists.tdwg.org [mailto:tdwg-content-bounces@lists.tdwg.org] On Behalf Of Tim Robertson [GBIF]
Sent: Tuesday, May 06, 2014 4:11 AM
To: Hilmar Lapp; Roderic Page; tdwg-content@lists.tdwg.org; "Markus Döring (GBIF)"; tomc@cs.uoregon.edu
Subject: Re: [tdwg-content] delimiter characters for concatenated IDs
Hi all,
Supposing GBIF or some other body were interested in offering such a central service as proposed in this thread. Can we articulate what we envisage would be the process?
a) client has a specimen record they wish to stamp with an identifier
b) client requests DOI (or other format) from the issuing service and provides the minimum metadata in a DwC-esque profile, potentially with a preferred suffix
c) service provides identifier, and client stores this along with their digital record
d) from this point on, the DOI identifies the record
If so, what would happen on resolution? Does the client provide the target URL during minting which will be the redirection target on resolution? Does the service have to monitor the availability and return the cached copy on outage?
In such a model, effectively we would have a central specimen registration service where data owners push individual specimen records. Is that something we envisage the community would accept? Presumably the minimum metadata would include things like dwc:scientificName - would someone register a DOI pushing that for specimens of a new species before they have published on the name?
This model will not in itself stop duplicate IDs. The scientists assembling datasets of specimens referenced in a paper might submit those references specimens for DOIs, while the original specimen curators might also submit the same records - thus the specimen is identified twice. Which piece of the infrastructure would capture that relationship?
What seems most important to me when I think this through is that the identifier needs to be minted as early on as possible in the record life - before it is shared with others. Which leads us back to the question of whether we envisage people adopting a model where they effectively submit their record data in order to get an identifier. If not, at least if we got stable IDs on records in whatever form, we can manage the resolvability bit later, and identify duplicates.
It would be interesting to hear how others imagine such a service operating.
Thanks,
Tim
On 06 May 2014, at 15:34, Hilmar Lapp <hlapp@nescent.org> wrote:
Every registration agency has its own set of standard metadata which members register for every DOI, but the content-negotiation strategy does allow for a richer metadata response. By default it is the registration agency's resolver that responds with RDF (and thus only with the metadata it knows of), but members (the entities registering DOIs) can register their own content-negotiation resolver, which would allow them to return richer metadata. We have, for example, considered doing this for Dryad (http://datadryad.org), but it hasn't risen to high-enough priority yet.
Hence, if GBIF were to register DOIs for specimens through DataCite (rather than being its own RA), then GBIF could still operate its own resolver for returning DwC metadata for RDF queries.
That doesn't mean there couldn't still be good arguments for GBIF serving as a RA.
-hilmar
On Tue, May 6, 2014 at 5:53 AM, Roderic Page <Roderic.Page@glasgow.ac.uk> wrote:
Hi Steve,
My understanding is that the non-HTML content is decided at the level of registration agency. For a bibliographic DOI registered with CrossRef, the HTML redirect goes to whatever the publisher provides CrossRef (e.g., the article landing page), other content (including RDF) is served by CrossRef based on the metadata they hold for each article. Likewise, DataCite will serve metadata based on what they have. Hence, metadata from CrossRef and DatacIte look rather different.
So, this is something that would need to be worked out at the level of registration agency (see http://www.crossref.org/CrossTech/2010/03/dois_and_linked_data_some_conc.html and http://crosstech.crossref.org/2011/04/content_negotiation_for_crossr.html for background).
Hence, if GBIF were to be a DOI registration agency they could serve Darwin Core RDF (and JSON and whoever else they want). This is a strong argument for GBIF doing this, rather than using DataCite (which serves very generic metadata).
Regards
Rod
On 6 May 2014, at 01:42, Steve Baskauf <steve.baskauf@vanderbilt.edu> wrote:
I'm a big fan of not reinventing the wheel, and as such find the idea of using DOIs appealing. I think they pretty much follow all of the "rules" set out in the TDWG GUID Applicability Standard. They also play nicely in the Linked Data universe in their HTTP URI form, i.e. they redirect to HTML or RDF depending on the request header.
But I have a question for someone who understands how DOIs work better than I do. The HTML representation seems to arise by redirection to whatever is the current web page for the resource. You can see this by pasting this DOI for a specimen into a browser: http://dx.doi.org/10.7299/X7VQ32SJ which redirects to http://arctos.database.museum/guid/UAM:Ento:230092 when HTML is requested by a client. However, when the client requests RDF, one gets redirected to a DataCite metadata page: http://data.datacite.org/10.7299/X7VQ32SJ . Can the creator of the DOI redirect to any desired URI for the RDF?
The resulting RDF metadata doesn't have any of the kind useful information about the specimen that you get on the web page but rather looks like what you would expect for a publication (creator, publisher, date, etc.):
<http://dx.doi.org/10.7299/X7VQ32SJ>
<http://purl.org/dc/terms/creator>
"Derek S. Sikes" ;
<http://purl.org/dc/terms/date>
"2004" ;
<http://purl.org/dc/terms/identifier>
"10.7299/X7VQ32SJ" ;
<http://purl.org/dc/terms/publisher>
"University of Alaska Museum" ;
<http://purl.org/dc/terms/title>
"UAM:Ento:230092 - Grylloblatta campodeiformis" ;
<http://www.w3.org/2002/07/owl#sameAs>
"info:doi/10.7299/X7VQ32SJ" , "doi:10.7299/X7VQ32SJ" .
Can one control what kinds of metadata are provided in "DataCite's metadata"? Assuming that we get our act together and adopt an RDF guide for Darwin Core, it would be nice for the RDF metadata to look more like the description of a specimen and less like the description of a book. But maybe that's just a function of where the data provider choses to redirect RDF requests.
Steve
John Deck wrote:
+1 on DOIs, and on ARKS (see: https://wiki.ucop.edu/display/Curation/ARK ), and also i'll mention IGSN:'s (see http://www.geosamples.org/) IGSN: is rapidly gaining traction for geo-samples. I don't know of anyone using them for bio-samples but they offer many features that we've been asking for as well. What our community considers a sample (or observation) is diverse enough that multiple ID systems are probably inevitable and perhaps even warranted.
Whatever the ID system, the data providers (museums, field researchers, labs, etc..) must adopt that identifier and use it whenever linking to downstream sequence, image, and sub-sampling repository agencies. This is great to say this in theory but difficult to do in reality because the decision to adopt long term and stable identifiers is often an institutional one, and the technology is still new and argued about, in particular, on this fine list. Further, those agencies that receive data associated with a GUID must honor that source GUID when passing to consumers and other aggregators, who must also have some level of confidence in the source GUIDs as well. Thus, a primary issue that we're confronted with here is trust.
Having Hilmar's hackathon support several possible GUID schemes (each with their own long term persistence strategy), and sponsored by a well known global institution affiliated with biodiversity informatics that could offer technical guidance to data providers, good name branding, and the nuts and bolts expertise to demonstrate good shepherding of source GUIDs through a data aggregation chain would be ideal. I nominate GBIF :)
John Deck
On Mon, May 5, 2014 at 1:09 PM, Roderic Page <r.page@bio.gla.ac.uk> wrote:
Hi Markus,
I have three use cases that
1. Linking sequences in GenBank to voucher specimens. Lots of voucher specimens are listed in GenBank but not linked to digital records for those specimens. These links are useful in two directions, one is to link GBIF to genomic data, the second is to enhance data in both databases, see http://iphylo.blogspot.co.uk/2012/02/linking-gbif-and-genbank.html (e.g., by adding missing georeferencing that is available in one database but not the other).
2. Linking to specimens cited in the literature. I’ve done some work on this in BioStor, see http://iphylo.blogspot.co.uk/2012/02/linking-gbif-and-biodiversity-heritage.html One immediate benefit of this is that GBIF could display the scientific literature associated with a specimen, so we get access to the evidence supporting identification, georeferencing, etc.
3. Citation metrics for collections, see http://iphylo.blogspot.co.uk/2013/05/the-impact-of-museum-collections-one.html and http://iphylo.blogspot.co.uk/2012/02/gbif-specimens-in-biostor-who-are-top.html Based on citation sod specimens in the literature, and in databases such as GenBank (i.e., basically combining 1 + 2 above) we can demonstrate the value of a collection.
All of these use cases depend on GBIF occurenceIds remaining stable, I have often ranted on iPHylo when this doesn’t happen: http://iphylo.blogspot.co.uk/2012/07/dear-gbif-please-stop-changing.html
Regards
Rod
On 5 May 2014, at 20:51, Markus Döring <mdoering@gbif.org> wrote:
Hi Rod,
I agree GBIF has troubles to keep identifiers stable for *some* records, but in general we do a much better job than the original publishers in the first place. We try hard to keep GBIF ids stable even if publishers change collection codes, registered datasets twice or do other things to break a simple automated way of mapping source records to existing GBIF ids. Also the stable identifier in GBIF never has been the URL, but it is the local GBIF integer alone. The GBIF services that consume those ids have changed over the years, but its pretty trivial to adjust if you use the GBIF ids instead of the URLs. If there is a clear need to have stable URLs instead I am sure we can get that working easily.
The two real issues for GBIF are a) duplicates and b) records with varying local identifiers of any sort (triplet, occurrenceID or whatever else).
When it comes to the varying source identifiers I always liked the idea of flagging those records and datasets as unstable, so it is obvious to users. This is not a 100% safe, but most terrible datasets change all of their ids and that is easily detectable.
Also with a service like that it would become more obvious to publishers how important stable source ids are.
Before jumping on DOIs as the next big thing I would really like to understand what needs the community has around specimen ids.
Gabi clearly has a very real use case, are there others we know about?
Markus
On 05 May 2014, at 21:05, Roderic Page <r.page@bio.gla.ac.uk> wrote:
Hi Hilmar,
I’m not arguing that we shouldn’t build a resolver (I have one that I use, Rich has mentioned he’s got one, Markus has one at GBIF, etc.).
Nor do I think we should wait for institutional and social commitment (because then we’d never get anything done).
But I do think it would be useful to think it through. For example, it’s easy to create a URL for a specimen. Easy peasy. OK, how do I discover that URL? How do I discover these for all specimens? Sounds like I need a centralised discover service like you’e described.
How do I handle changes in those URLs? I built a specimen code to GBIF resolver for BioStor so that I could link to specimens, GBIF changed lots of those URLs, all my work was undone, boy does GBIF suck sometimes. For example, if I map codes to URLs, I need to handle cases when they change.
If URLs can change, is there a way to defend against that (this is one reason for DOIs, or other methods of indirection, such as PURLs).
If providers change, will the URLs change? Is there a way to defend against that (again, DOIs handle this nicely by virtue of (a) indirection, and (b) lack of branding).
How can I encourage people to use the specimen service? What can I do to make them think it will persist? Can I convince academic publishers to trust it enough to link to it in articles? What’s the pitch to Pensoft, to Magnolai Press, to Springer and Elsevier?
Is there some way to make the service itself become trusted? For example if I look at a journal and see that it has DOIs issued by CrossRef, I take that journal more seriously than if it’s just got simple URLs. I know that papers in that journal will be linked into the citation network, I also know that there is a backup plan if the journal goes under (because you need that to have DOIs in CrossRef). Likewise, I think Figshare got a big boost when it stared minting DOIs (wow, a DOI, I know DOIs, you mean I can now cite stuff I’ve uploaded there?).
How can museums and herbaria be persuaded to keep their identifiers stable? What incentives can we provide (e.g., citation metrics for collections)? What system would enable us to do this? What about tracing funding (e.g., the NSF paid for these n papers, and they cite these y specimens, from these z collections, so science paid for by the NSF requires these collections to exist).
I guess I’m arguing that we should think all this through, because a specimen code to specimen URL is a small piece of the puzzle. Now, I’m desperately trying not to simply say what I think is blindingly obvious here (put DOIs on specimens, add metadata to specimen and specimen citation services, and we are done), but I think if we sit back and look at where we want to be, this is exactly what we need (or something functionally equivalent). Until we see the bigger picture, we will be stuck in amateur hour.
Take a look at:
Isn’t this the kind of stuff we’d like to do? If so, let’s work out what’s needed and make it happen.
In short, I think we constantly solve an immediate problem in the quickest way we know how, without thinking it through. I’d argue that if we think about the bigger picture (what do we want to be able to, what are the questions we want to be able to ask) then things become clearer. This is independent of getting everyone’s agreement (but it would help if we made their agreement seem a no brainer by providing solutions to things that cause them pain).
Regards
Rod
On 5 May 2014, at 19:14, Hilmar Lapp <hlapp@nescent.org> wrote:
On Mon, May 5, 2014 at 1:29 PM, Roderic Page <r.page@bio.gla.ac.uk> wrote:
Contrary to Hilmar, there is more to this than simply a quick hackathon. Yes, a service that takes metadata and returns one or more identifiers is a good idea and easy to create (there will often be more than one because museum codes are not unique). But who maintains this service? Who maintains the identifiers? Who do I complain to if they break? How do we ensure that they persist when, say, a museum closes down, moves its collection, changes it’s web technology? Who provides the tools that add value to the identifiers? (there’s no point having them if they are not useful)
Jonathan Rees pointed this out to me too off-list. Just for the record, this isn't contrary but fully in line with what I was saying (or trying to say). Yes, I didn't elaborate that part, assuming, perhaps rather erroneously, that all this goes without saying, but I did mention that one part of this becoming a real solution has to be an institution with an in-scope cyberinfrastructure mandate that going in would make a commitment to sustain the resolver, including working with partners on the above slew of questions. The institution I gave was iDigBio; perhaps for some reason that would not be a good choice, but whether they are or not wasn't my point.
I will add one point to this, though. It seems to me that by continuing to argue that we can't go ahead with building a resolver that works (as far as technical requirements are concerned) before we haven't first fully addressed the institutional and social long-term sustainability commitment problem, we are and have been making this one big hairy problem that we can't make any practical pragmatic headway about, rather than breaking it down into parts, some of which (namely the primarily technical ones) are actually fairly straightforward to solve. As a result, to this day we don't have some solution that even though it's not very sustainable yet, at least proves to everyone how critical it is, and that the community can rally behind. Perhaps that's naďve, but I do think that once there's a solution the community rallies behind, ways to sustain it will be found.
-hilmar
--
Hilmar Lapp -:- informatics.nescent.org/wiki -:- lappland.io
_______________________________________________
tdwg-content mailing list
tdwg-content@lists.tdwg.org
http://lists.tdwg.org/mailman/listinfo/tdwg-content
--
John Deck
(541) 321-0689
--
Steven J. Baskauf, Ph.D., Senior Lecturer
Vanderbilt University Dept. of Biological Sciences
