DCC Workshop on Persistent Identifiers A Layered Model Decision Making Concerning Persistent Identif

1
DCC Workshop on Persistent IdentifiersA
Layered Model Decision Making Concerning
Persistent Identifiers

• Stuart Weibel
• Senior Research Scientist
• July 1, 2005

2
Open Univ

• How do you establish that you need a persistent
  identifier (as opposed to a non-persistent one)
• E.g. OU is creating a new course, with a very
  complex collection of courseware, and an equally
  complex workflow which manages all the parts
• At what stage do we assign persistent identifiers
  to all the bits?
• SW Is this a granularity issue?
• In part, but theres ownership issues

3
CETIS

• Need to be careful about advice, and who we give
  it to
• Wheres the line between what you do and dont
  need to care about
• At what point does persistence become important
• PB publish is the crucial word here

4
LSE

• One or many systems?
• Digitising a photo archive, use a separate system
  for that?
• Who assigns identifiers at the individual school
  level
• Managing identifier incrementation
• What about managing versions in a repository?
• Separate identifiers for distinct manifestations
  of the same article?

5
Nat Lib Scot

• Anyone with experience of Nat Lib of Aus scheme?

6
Strathclyde Univ CDLR

• Confession I assign identifiers
• And Ive changed them
• And changed their URLs
• But, Ive left redirects
• Is this ok?

7
Edin Univ Library

• How relate identifiers for a physical object to
  digital manifestations of that object
• Preexisting vs. internal institutional
  identifiers
• Public vs. private
• How implement identifiers within a METS document
• Enders Not a METS problem, an XML problem
  SICI is not an XML ID

8
Edin Univ Lib

• If http URIs are a given, should they be opaque
  (e.g. ISBNs) or semantic (e.g. http//www.w3.org/R
  EC-xml/)
• Do you know your own barcode?
• Semantic
• Human readable (within a designated community,
  broad (native speakers of English) to narrow
  (well-trained biomed librarians)
• vs
• Rule-based, so decodable given some additional
  info
• Opaque
• E.g. ARK, random string, with (by design) no
  structure
• E.g. LibCon(?) concatenation of year and
  accession number
• What am I trading off wrt to one choice or
  another?
• Are http URIs intrinsically semantic?

9
Enders

• If weve assigned identifiers, in a non-online
  repository, how can I get a persistent
  identifier?
• E.g. DSPACE
• AD DSPACE uses Handle
• Cetis Fedora uses Handle (?)
• Strathclyde Eprint.org does not offer any

10
CETIS

• Authoring and content tools should help out here
• SW Where in workflow should identifiers be
  assigned?

11
EDINA

• To what extent should we have an international
  non-territorial view of assigning identifiers?
• Or should we stay within the bounds of the nation
  state?

12
Glasgow Univ Lib

• If from my repository I want to refer to some
  other repositorys identifier, in a persistent
  way
• Can I do so responsibly independently of how
  theyve done this
• JK If they dont publish a name they take
  responsibility for, dont go there!
• Image archive example you can email a filename
  and get a temporary URI (ftp) back how can I
  write down into a formal workflow a reference to
  BigBear file name so-and-so
• So its a permanent identifier for a static
  digital resource, but I cant refer to it
  formally
• RR instance of general case of offline
  resources
• SW They need to be pressured to recognise the
  value of e.g. prepending http
• But they wont do that today, so Im going to do
  it, but that feels like breaking the rules
• ?? What about names for resources in GRID
  computing?
• ?? One option is Handles, recently released
  module for Globus
• ?? Surely allocating persistent http URIs
  would be overkill with 5M large images
• SW Publishing the naming structure is distinct
  from delivering all of them instantly
• tutti design-on-the-fly

13
SunCat

• Working on a catalogue of serials etc., heard
  about FRBR, etc., but inadequate for my needs, do
  I use the same identification scheme up and down
  the hierarchy
• AP important to distinguish between
  application-level questions and in-principle
  general questions
• PB Application-level are not ipso facto out of
  scope
• And whats an expression?
• SW Out of scope

14
UKOLN

• Candidate functional requirements for persistent
  identifiers
• Compare for equality
• Retrieve from it
• Do we get more specific, e.g. ARK?
• Ding an sich JK or suitable surrogate
• HST Demurs
• Description
• Cover the case where the identifier no longer
  identifies anything?
• Change history
• PB Theres a wealth of relevant experience
• Persistence commitment
• SW Demur no always desirable, sometimes not
  desirable
• Encode and distribute

15
NLScotland

• Where would we go in ISO to do this
• PB There is an ISO technical cttee
• RR Wrt NISO, I can feed back on where do we go
  Standards, recommendations, Best Practices notes

16
Open Univ

• Can we clarify whats been done wrt Premis data
  dictionary?
• Just published a new dictionary
• JK Wrt identifiers, says nothing, judged too
  controversial/out of scope

17
DCC

• Is there a relationship between the identifier
  and the findability of a resource?
• Given pervasive use of search engine
• SW, PB What are the roles in discover and
  location of service

18
British Library

• If we do use http URIs for identifiers, does the
  HTTP protocol provide any particular actions
  which make sense, e.g. PUT or POST?
• Should resolution be a determiner of existence
• SW redirection comes in here
• In theory, any 10/13 digit number can be an ISBN
• What does 404 not found mean?

19
Strathclyde

• Impact of search engines people arent using
  identifiers any more
• Should we question to assumption that identifiers
  are the entry point into access to resources

20
DCC/Hunter

• Whats the impact of the fact that the community
  is split into two parts
• Those with detailed knowledge of the identity
  system
• Those who just have requirements which identity
  contributes to solving
• Bookshop example
• PB Ironmonger example

21
Discussion candidates

• Semantic vs. opaque
• What should libraries do
• Is retrievable a necessary functional
  requirement
• What are we identifying
• What is the connection between the identifier and
  the think itself?

22
Seamus/DCC

• Can we do automatic assignment of identifiers to
  extracts from (multiple) databases
• RR About attaching metadata as you go
• As an author, a private person, how do I get an
  identifier for my work
• Overwriting is OK, always?
• Historical printing example feeds into thinking
  about multiple versions in the digital world
• PB Relevant term is provenance and the role
  of identifiers

23
UKOLN

• Are the functional requirements on identifiers
  across domains or classes of resources?

24
EDINA

• Is the distinction between persistence for
  record and persistence for reuse relevant to
  the design of identifiers?
• Archival vs. access?
• SW Difference in behaviour based on difference
  in object identified
• Seamus BNFL vs. BBC example

25
In the Information World

• We care about identifying resources
• Physical
• Virtual
• Conceptual
• Knowing you have what you think you have
• Comparing identity (referring to the same thing)
• Reference linking
• Managing intellectual (or physical) property

26
What do we want from Identifiers

• Authority
• Reliability
• Appropriate Functionality (resolution and other
  services)
• Persistence throughout the life cycle of the
  information object
• What are the business models to support
  identifiers?
• Not just a matter of money, but costs are part of
  the equation

27
The Identifier Layer Cake

• Identifiers come in many sizes, flavours, and
  colours what questions do we ask?

Social
Business
Policy
Technology
Functionality
The Web httpTCP/IPfuture infrastructure?
28
Functional Layer Operational characteristics of
Identifiers

• Is it globally unique? (easy)
• What is the means for matching persistence with
  the need?
• Can a given identifier be reassigned?
• Is it resolvable? To what?
• How does it behave? What applications
  recognize it and act on it appropriately?
• Is the name portion of the identifier opaque,
  or can it carry semantics?
• Do humans need to read and transcribe them?
• Do identifiers need to be matched to the
  characteristics of the assets they identify?

29
Information Assets have life cycles with
different characteristics

• Journal Articles
• Created
• Reviewed
• Pre-published Published
• Versioned
• Sold Resold
• Archived
• Cited
• Distributed in a variety of channels (appropriate
  copy problem)

• Concepts Terms
• Created and deprecated
• Versions
• Definitions
• Abstract (concepts) and instantiated (terms)
• Translations (for terms)
• Position in a hierarchy (ontology)
• Relations, linkages

30
Technical Layer

• What dependencies are assumed?
• http tcp/ip(bar codeRFID) scanners
• What is the nature of the systems (both software
  and social) that support assignment, maintenance,
  resolution of identifiers?
• Are servers centralized? federated? peer to peer?
• How is uniqueness assured?

31
Policy Layer

• Who has the right to assign or distribute
  Identifiers?
• Who has the right to resolve them or offer
  serves against them?
• What are appropriate assets for which identifiers
  can be assigned, and at what granularity?
• Can identifiers be recycled?
• Can ID-Asset bindings be changed?
• Is there supporting metadata, and if so, is it
  public, private, or indeterminate?
• Is there a governance model?

32
Business model layer

• Who pays the cost?
• How, and how much?
• Who decides (see governance model)?
• The problem with identifier business models
• Those who accrue the value are often not the same
  as those who bear the costs
• You probably cant collect revenue for resolution

33
Social Layer

• The only guarantee of the usefulness and
  persistence of identifier systems is the
  commitment of the organizations which assign,
  manage, and resolve identifiers
• Who do you trust?
• Governments?
• Cultural heritage institutions?
• Commercial entities?
• Non-profit consortia?
• We trust different agencies for different
  purposes at different times

34
Terminology Identifiers

• Global, persistent identifiers that reflect the
  functional characteristics of webulated
  controlled vocabularies can help us remove
  boundaries between and among communities and
  disciplines.
• Problem Identify these functional requirements
  and tailor identifier systems to meet them.

35
Identifiers for Concepts

• How do you use terminology in the Web World?
• The Semantic Web is about semantics exchanging
  tokens of meaning between machines
• Identifiers are a fundamental part of this.

36
Concepts can be expressed in language independent
ways (even if imperfectly)

• Vietnamese War, 1961-1975
  DDC/22/eng//959.7043
• (English language version of DDC 22)
•  
• American War, 1961-1975 DDC/22/vie//959.7043
  
• (Vietnamese language version of DDC 22,)

37
Boundary-Free Community Terminologies

• Controlled Vocabularies have been with us for a
  long time
• Hypothesis there are specific functional
  requirements that terminologies should embody in
  order to be useful in the realization of the
  Semantic Web