Finally I found some time to go through this lively thread. I hope my post is not already outdated by someone elses ;)
Apart from the what gets an URI discussion there have been some people expressing their doubts about LSIDs. As I have had a number of discussions lately with people doubting that LSIDs are good for our purposes, I would really like to question the TDWG decision to go with LSIDs and start yet another comparison of plain http paired with redirection, content negotiation and guidelines for using URLs. I strongly feel that we should avoid new protocol schemes if we do not have *very* good reasons. I will use the term URL for now to refer to any http based identification scheme, if its PURLs, our own system or something else.
The LSID specification already tells us how to deal with persistent identifiers. It is an agreement that we would have to make for URLs. As the "what gets an URI" confusion has shown those guidelines are needed in any case, no matter if we take up LSIDs or not. Even LSIDs can be used with or without versioning and a lot depends on agreements in regard to the RDF behind it. So essentially we will have to come up with our own best practices anyway.
LSID and HTTP both are based on DNS to guarantee global uniqueness and even more important to resolve them. They both derive their persistence from the promise of the service provider that the domain name is kept forever and a server is running. If the domain is lost in 50 years *both* systems are broken.
LSIDs and the semantic web dont play nicely together per se, cause the semweb de facto requires plain http. From what I've read the suggestion is to use an LSID proxy that maps URLs into LSIDs. The problem then is that all RDF statements must use the proxy URL instead of the real LSID (otherwise you/a resoner doesn't know that the statement about the LSID and the statement about the proxyURL are about the identical resource) so essentially noone is using the LSIDs, they are just kept as an additional "persistent" ID. To overcome this problem and to be able to use both, the LSID or the proxy URL, it is suggested to use an owl:sameAs statement within the LSID metadata to link the proxy URL with the LSID. So applications can use this to understand we are talking about the same thing. This gets pretty complex already and I would be surprised if there are many applications out there that understand this.
Why not apply the owl:sameAs trick to URLs once we find that http is dead (just in case we can't do a global search-and-replace)? We could stay with simple URLs now, write simple software fast and get into the complex mess at a much later stage when we know we really need to - and not already from the start.
A very often raised requirement for the technology is also that it should last for hundreds of years. I doubt anyone can predict in that time period. But a very good reason to go with http is that there is a *lot* of data bound to them and if the world decides there is something better than http, there will be many tools to migrate your data. I feel much more safe trusting the entire web community than eventually getting out of the LSID trap by ourselves.
Imagine if all the different research communities decide to use their own resource identification scheme, how bad will data integration get? We have to deal already with DOIs, but imagine chemists, geologists, meteorologists, physicists would all choose their own scheme, just as we are about to issue life science identifiers? Non- http URIs put barriers up for adoption to other communities, so I am confident that our LSIDs will be referenced much much less than URLs. I can see already all those proxy URLs in genebank and alike, not the LSIDs.
And finally yet another link to some good discussion in the W3C semweb lifescience list: http://lists.w3.org/Archives/Public/public-semweb-lifesci/2005Mar/ 0004.html
-- Markus
On 06.06.2007, at 09:21, Dave Vieglais wrote:
This discussion has been very interesting reading, and though I agree with Donald's comments, I find myself coming to a different conclusion, leaning towards HTTP URIs as a preferable scheme. The reasons are simple - HTTP has been around for a long time, it is widely implemented, and mechanisms for implementing robust services with that protocol are pretty well sorted out - and really there is nothing to stop implementation of the same functionality exhibited by LSIDs using HTTP. As Rod has pointed out, http is widely used for referencing entities within a semantic web type of context, and it seems foolish to ignore the momentum in those technologies as they provide a great deal of the desired functionality for interoperability and interchange of our data. As a result my preference is towards the use of http, primarily because my intents are to integrate data from a much broader community. In the end though, it doesn't really matter which scheme is adopted by TDWG - we will build http resolvers regardless, since they will be necessary for reasons of convenience in order to utilize LSIDs in all but specific, custom built applications.
However, regardless of the scheme used to implement the GUIDs used by this community, it is critical that the identifiers are persistent and useful beyond the lives of whatever services are constructed to resolve them. This implies some provenance information may need to be captured, and I would argue that the use of DNS alone for handling server changes as utilized by LSIDs may be insufficient. The only benefit provided by DNS in this context is that it is acting as a single source of authority for directing how to locate something (in this case an ip address). What I suspect is really required is a more robust, and richer mechanism for discovering and recording provenance. The ideal would be a large, replicated, and distributed data store with a single service point which provided people and systems with a one-stop shop for discovering provenance for a GUID. Then if an particular GUID could not be directly resolved, the global provenance store could be consulted and the resulting information providing a pointer (or perhaps a series of pointers) indicating how the guid can now be resolved.
By creating such provenance records and persisting them with as much care as the data, it seems that a system with stability beyond the vagaries of the internet could reasonably be constructed.
regards, Dave V.
On Jun 6, 2007, at 00:46, Donald Hobern wrote:
Yesterday was a vacation here in Denmark - otherwise I'd have responded a little earlier, but I'm glad to see all the comments from others. I thoroughly agree with Kevin, Jason, Rich and Anna. No one here believes that any particular solution is going to be perfect. Our biggest need is consensus and the readiness to get going with a workable solution.
I do recognise the strength of Rod's arguments. Indeed, if I were building some system for integrating data using semantic web technologies, and my only concern was ensuring the efficiency of synchronous connections now, I am sure I would adopt HTTP URIs for the purpose. However I remain convinced (as I've stated before) that the needs of this community do subtly shift the balance in another direction. We are interested in maintaining long-term connections between our objects and have a perspective which goes back hundreds of years. This at least should give us pause over whether we want our specimens to be referenced using identifiers so firmly tied to the Internet of today. More importantly, one of the key drivers right at the beginning of TDWG's consideration of GUIDs was that the community had plenty of experience of URL rot and didn't want to rely on everyone maintaining stable virtual directories on their web servers to preserve the integrity of object identifiers.
Both LSIDs and HTTP URIs could be made to work for us. Both are totally reliant on good practice on the part of data owners. Personally I believe our chances of getting the community to consider, define and apply such practices are enhanced by the identifier technology being something a little more different and distinct than just a "special URL".
Thanks,
Donald
Donald Hobern (dhobern@gbif.org) Deputy Director for Informatics Global Biodiversity Information Facility Secretariat Universitetsparken 15, DK-2100 Copenhagen, Denmark Tel: +45-35321483 Mobile: +45-28751483 Fax: +45-35321480
On Jun 6, 2007, at 12:51 AM, Kevin Richards wrote:
I agree with Jason. It is not the GUID that is the cause of all the problems here - THERE IS NOTHING WRONG WITH LSIDS - we just need to move on and start using them in our own context (or any other suitable GUID - LSIDs are only the recommended GUID, NOT the only premissable GUID).
If it all falls to pieces later on we could just do a search and replace to change all our GUIDs to some other scheme (to quote Bob, just serious).
I agree, it is the RDF/metadata/ontologies that are the key to getting things working well.
Kevin
"Jason Best" jbest@brit.org 06/06/07 8:39 AM >>>
Rod, I've only had a chance to quickly skim the documents you reference, but it seems to me that the alternatives to LSIDs don't necessarily make the issues with which we are wrestling go away. We still need to decide WHAT a URI references - is it the metadata, the physical object etc? URIs don't explicitly require persistance, while LSIDs do so I see that as a positive for adopting a standard GUID that is explicit in that regard. I think the TDWG effort to spec an HTTP proxy for LSIDs makes it clear that the technical hurdles of implementing an LSID resolver (SVR records, new protocol, client limitations etc) are a bit cumbersome, but I don't think the underlying concept is fatally flawed. In reading these discussions, I'm starting to believe/ understand that RDF may hold the key, regardless of the GUID that is implemented. Now I have to go read up more on RDF to see if my new-found belief has merit! ;)
Jason
From: Roderic Page [mailto:r.page@bio.gla.ac.uk] Sent: Tuesday, June 05, 2007 2:10 PM To: Chuck Miller Cc: Bob Morris; Kevin Richards; tdwg-guid@lists.tdwg.org; WEITZMAN@si.edu; Jason Best Subject: Re: [tdwg-guid] First step in implementing LSIDs?[Scanned]
Maybe it's time to bite the bullet and consider the elephant in the room -- LSIDs might not be what we want. Markus Döring sent some nice references to the list in April, which I've repeated below, there is also http://dx.doi.org/10.1109/MIS.2006.62 .
I think the LSID debate is throwing up issues which have been addressed elsewhere (e.g., identifiers for physical things versus digital records), and some would argue have been solved to at least some people's satisfaction.
LSIDs got us thinking about RDF, which is great. But otherwise I think they are making things more complicated than they need to be. I think this community is running a grave risk of committing to a technology that nobody else takes that seriously (hell, even the http://lsid.sourceforge.net/ web site is broken).
The references posted by Markus Döring were:
(1) http://www.dfki.uni-kl.de/dfkidok/publications/TM/07/01/ tm-07-01.pdf "Cool URIs for the Semantic Web" by Leo Sauermann DFKI GmbH, Richard Cyganiak Freie Universität Berlin (D2R author), Max Völkel FZI Karlsruhe The authors of this document come from the semantic web community and discuss what kind of URIs should be used for RDF resources.
(2) http://www.w3.org/2001/tag/doc/URNsAndRegistries-50 This one here is written by the W3C and addresses the questions "When should URNs or URIs with novel URI schemes be used to name information resources for the Web?" The answers given are "Rarely if ever" and "Probably not". Common arguments in favor of such novel naming schemas are examined, and their properties compared with those of the existing http: URI scheme.
Regards
Rod
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++ WARNING: This email and any attachments may be confidential and/or privileged. They are intended for the addressee only and are not to be read, used, copied or disseminated by anyone receiving them in error. If you are not the intended recipient, please notify the sender by return email and delete this message and any attachments.
The views expressed in this email are those of the sender and do not necessarily reflect the official views of Landcare Research.
Landcare Research http://www.landcareresearch.co.nz ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++
tdwg-guid mailing list tdwg-guid@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-guid
tdwg-guid mailing list tdwg-guid@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-guid
tdwg-guid mailing list tdwg-guid@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-guid