Manfred Thaller Universität zu Köln [email protected] Preserving for 2016, 2106, 3006 Or Is there a life for an object outside a digital library?

Manfred ThallerUniversität zu Köln

[email protected]

Preserving for 2016, 2106, 3006Or

Is there a life for an object outside a digital library?

DELOS International Summer School 2006San Miniato

4-9 June 2006

DELOS Int SS2006, San Miniato, © Manfred Thaller, Universität zu Köln 2

A persistent object


A persistent object

•Authenticity

•Integrity

•Metadata

•Context

•Easily usable


A persistent object

•Authenticity

•Integrity

•Metadata

•Context

•Easily usable

•Discussable


A persistent object

•Authenticity

•Integrity

•Metadata

•Context

•Easily usable

•Discussable

•No


A persistent object

•Authenticity

•Integrity

•Metadata

•Context

•Easily usable

•Discussable

•No

•No


A persistent object

•Authenticity

•Integrity

•Metadata

•Context

•Easily usable

•Discussable

•No

•No

•No


A persistent object

•Authenticity

•Integrity

•Metadata

•Context

•Easily usable

•Discussable

•No

•No

•No

•1799 - 1821


Persistent until: 2016

•No major breakdown of civil society.

•'Library' system continues to function without serious interruption.

•No fundamental change in underlying technology.

•No major 'holes' in the relevant WWW.



Assumption therefore:Persistency is a function of the overall system.










•'Library' changes functional assumptions without serious interruption of service.







•Fundamental changes in underlying technology.






•Fundamental changes in underlying technology.

•Major 'holes' in the relevant WWW.



Assumption: Persistent storage media "around the corner". (Holographic storage, storage crystals.)

Question: Can a digital object be revived in 2106, if library does not care for it after 2016?



Why not?

•Bit stream deterioration.•Authenticity not guaranteed.•Meta data get lost.•Context gets lost.


Bit stream deterioration

An Image filebefore ….



... and afterone byte ischanged.



... and afterone byte ischanged.

Undetectableby software.



Sketch of a technical solution.

Underlying assumption:bit stream deterioration becomes less of aproblem, if "files" are designed for persistency.



Proposal 1: Measure File Robustness

Proposed metric: A file is m / n robust, if you can change m arbitrarily selected bytes ofthe stored data without affecting more than n bytes of the payload bytesof the file. Background terminology: Any file format can be described as consisting of a processing dictionary(roughly: technical metadata) and a payload, which represents theinformation presented to the user. Proposed implementation: Apply at least one thousand / one million timesrandom change to n randomly selected byte and get mean number ofaffected bytes..



Proposal 2: Measure Error awareness

Proposed metric: A file / file reader is n error aware, if you can change at the mostn arbitrarily selected bytes of the stored data without it becomingdetected during every attempt at reading.

Background terminology: Any file format which has predicted lenghts for each reading operationplus some additional info on the result of the reading operation has thisproperty to some degree.

Proposed implementation: Experiment to understand the situation better.



Proposal 3: Improve relevant file qualities - Hardening

Proposed metric: A file is n hardened, if it contains n synchronized redundant copies of theprocessing dictionary.

Background terminology: Two chunks of data are synchronized, if a processing environmentguarantees, that they are always changed in parallel.

Proposed implementation: Create TIFF / PNG writers / readers, whichsignal by additional tag / chunk the further copies of the processingdictionary.


Bit stream deteriorationProposal 4: Improve processing capabilities – Self repairing

Definition: A file is self repairing, if a reader is able to recover, after discoveringthat internal data are missing.

Example: PDF files tolerate modest distortions, as they are able to identifythe beginning of major sections within the file.


Authenticity not guaranteed.

Problem:While paper has physical properties, which can be evaluated,digital documents do not.

Solution:Add digital signatures, recognisable by dedicated software,registered with suitable authority.




Solution:Add digital signatures, recognisable by dedicated software,registered with suitable authority.

Violates assumption of change of framework.




Proposal:Insert fingerprint of institution (potentially individual PC) Implicitly into every file generated.



Binary file sealing:

1) Modify payload to provide parity within small formin byte stream.

2) Select arbitrary start address within payload.

3) Build path of parity forms.




Proposal:Insert fingerprint of institution (potentially individual PC) Implicitly into every file generated.

Problem: Incompatible with logic of storing text as XML.


Meta data get lost.

"Metadata" and data are stored separately in currentInformation system designs.

Take an image data base as example.


Meta data get lost.


Meta data get lost.


Meta data get lost.

"thumbs.db, but more so"


Meta data get lost.

MA thesis Jan Schnasse: http://lehre.hki.uni-koeln.de/~schnasse/ediod/; [email protected]


Meta data get lost.



Meta data get lost.



Meta data get lost.



Meta data get lost.



Context gets lost ...

More on properties of preservation aware files: Localized

Definition: A file is localized, if a reader can process it without accessing aremote server.

Counterexample: Virtually all XML-based standards of the DL community assume,that a program processing the file has access to a fully operational web,preferably in the structure of 2005, and / or the functioning ofauthorities like a URN resolution mechanism.

Solution: Snapshot of refered components.



More on properties of preservation aware files: Autonomous

Definition: A file is autonomous, if a reader can process it without accessinganother file.

Counterexample: A PDF is usually not autonomous in the strict sense, as it assumes thatfont information is available. Could be discussed at length, as it comesdown to the question, which resources are defined as part of theprocessing environment.

Solution: "Discuss at length".



More on properties of preservation aware files: Selfdocumenting

Definition: A file is selfdocumenting, if it contains as part of the processingdictionary a complete set of metadata.

Solution: Register appropriate tags / chunks with the TIFF / PNG authorities.



More on properties of preservation aware files: Preservation encapsulated

Definition: A file is preservation encapsulated, if it starts with a preservationheader, acting as processing dictionary for a subset of the capabilitiesdefined from "hardening" to "self documenting", and continues with astandard file of a recognized standard format.

Solution: Well, if we can register URNs, why should we not be able to maintain anencapsulation format?



More on properties of preservation aware files: Preservation encapsulated

Definition: A file is preservation encapsulated, if it starts with a preservationheader, acting as processing dictionary for a subset of the capabilitiesdefined from "hardening" to "self documenting", and continues with astandard file of a recognized standard format.

Solution: Well, if we can register URNs, why should we not be able to maintain anencapsulation format?

Registry violates autonomy, however.


Group Exercise

Background:(1) Assume that sometime betweennow and 2106 your institution gets into a seriousfinancial crisis. ("Serious" are e.g. budgetcuts to, not by, 5 % of previous year.)(2) Your digital data have to survivefor 30 years on their own under extremely badconditions, where random deletions area certainty.


Group Exercise

Planning for that now:(1)What would be the most sensible autonomousunit to divide your holdings into, making eachunit able to survive on its own.(2) Units which are to small are disastrouslyredundant; units which are to big, are disastrouslyvulnerable.


Group Exercise

Notice to people with IT training:The usual warnings against redundancy indata base design are related to "living" databases, avoiding "anomalies", which can notoccur in longterm storage.(If you do not understand the above, youmay safely ignore it.)









•Major breakdown of civil society.







•'Library' system seriously interrupted.







•'n' fundamental changes in underlying technology.






•'n' fundamental changes in underlying technology.

•WWW completely replaced by another type of connectivity



Any chance at all?



Any chance at all?

No real answer, but some stuff for thinking.



Is this Information ?












1. Recognizing information

2. Technological assumptions

3. Cultural assumptions

4. Processing assumptions



1. Announcement headers introducing hierarchically into the assumptions?

2. Preservation encapsulation with different horizons?


Group Exercise

(1) List all the assumptions, the material which isstored in printed form in your institution makesof the background knowledge if the reader.

(2) Design an "announcement header" for suchinformation.

Manfred Thaller Universität zu Köln [email protected] Preserving for 2016, 2106, 3006 Or Is there a life for an object outside a digital library?

Documents

san miniato

delos int ss2006

usable slide

usable discussable slide

major holes

persistent storage media

digital object

fundamental changes