InterPARES3 FileViewers
InterPARES3 FileViewers
Report
Document Control
Version history
1.2 2016-02-16 R. Preston Minor content and copy edits for public version.
Table of Contents
B. Methods ............................................................................................................................................... 5
Product Selection................................................................................................................................ 5
Software Comprehension .................................................................................................................. 6
Test Environment Set-Up .................................................................................................................. 6
File Formats, File Properties and Characteristics, and Files Selection ....................................... 6
Data Collection ................................................................................................................................... 8
C. Results .................................................................................................................................................. 8
Product Selection Results .................................................................................................................. 8
Software Comprehension Results .................................................................................................... 8
Test Environment Set-Up Results .................................................................................................. 10
File Formats, File Properties and Characteristics, and Files Selection Results ....................... 11
Data Results ...................................................................................................................................... 12
D. Conclusion ......................................................................................................................................... 15
E. Acknowledgements .......................................................................................................................... 19
F. References ......................................................................................................................................... 20
During the InterPARES 3 TEAM Canada case study on preparing shared drive files for
migration into an electronic content management (ECM) system (Rogers et al., 2010), the co-
investigators at the City of Surrey identified and adopted a number of utility applications to
expedite their work. These utility applications included: a disk space manager, used to collect
drive statistics, analyze file formats, create historical profiles, and facilitate metadata discovery; a
file manager, used to apply a unique identifier and rename records; a duplication finder, used to
identify and remove duplicates based on byte-by-byte comparison; a format identifier, used to
identify and resolve missing file extensions; and a empty folder identifier, used to count and
(initially) remove empty folders. The use of these utility applications is described in the Shared
Drive Migration Toolkit (Enns and Badesha, 2011). During the course of the migration work,
over 285 file formats were identified and appraised, however, following the business, technical,
and records analysis described in the toolkit, only 47 file formats were confirmed as records
suitable for migration, and only two of these file formats were found to be obsolete (unpublished
data). These two file formats (.ptn and .dwt) represented only 18 files out of 98,000 appraised for
migration. The balance of the file formats confirmed as records could be opened, dependent on
the availability of the native application.
During the shared drive migration and ECM implementation, the City of Surrey team
made a decision not to convert any files to long-term preservation formats. Beyond the
constraints of time and available resources, the shared drive environment was ill-suited to linking
the native and preservation files, and managing the associated metadata. By contrast, specific
features in the ECM system could be leveraged to meet these ends post-migration. Additionally,
in an active record system, there were difficulties in identifying the point-in-time at which a file
should be put on a migration pathway. Although some record series could be converted to a
preservation format on a time-based schedule, other record series reflected event-based activities
where secondary files suitable for migration were generated for other purposes during the record
lifecycle and could be co-opted for preservation. Finally, given the small number of formats
identified as being patently obsolete (i.e., two), there did not appear to be a pressing need to
begin bulk conversions at this time. In fact, the co-investigators wondered whether the question
of file conversion might be postponed indefinitely, given the ability to open the native files. In
relation to this, the ECM system purchased by the City of Surrey included a file viewer that
allowed users to open and annotate specialty drawing files where they did not have the native
application loaded on their computer. Although subsequent testing revealed that the viewer
module was not well integrated to the ECM system and it was not adopted, the idea that a file
viewer might extend the life of a file format was appealing. As a secondary consideration in
favour of investigating file viewers, the co-investigators found that during pre-migration file
renaming activities, opening files to validate contents was a time-consuming activity, since only
a few applications could be kept open on the task bar, and time was spent waiting for
applications to open or load, and in flipping between native and utility applications. A file viewer
that would enable viewing of multiple formats from a single point appeared to be an avenue
worth pursuing.
Seeking to investigate file viewer further, the co-investigators worked with InterPARES 3
to formulate research questions, and four areas of interest were identified: how do file viewers
work; what software is available for use; how accurately do file viewers render files; and what
role might file viewers play in digital preservation. Over the course of a year, these questions
were examined by the co-investigators and discussed at InterPARES 3 bi-annual workshops.
From the outset, the research questions represented an unexplored area of interest, since little
was available on the topic, and none of the case study participants or researchers used file
viewers in their work at that time, with the exception of a research assistant who had used a file
viewer while working at the Washington State Archives.
Literature Review
A number of articles mentioning file viewers are found in software and computer
engineering journals, primarily with respect to the role of file viewers in software design. For
example, an article on a product called GroupKit mentions a file viewer in the context of enabling
users’ views of text documents in a Unix conferencing environment (Roseman and Greenberg
1995, p.6). Similarly, two articles mention file viewers in the context of Unix programming, and
list other types of viewers: a directory viewer, an error viewer, an execution viewer, a software
landscape viewer, and an interface viewer (Manoridis et al., 1993, pp.16, 18) and a project viewer
and a graph viewer (Anderson and Teitelbaum, 2001, p.3). Interestingly, file viewers are one of a
Preservation and Long-term Access through Networked Services) and national archives (e.g.,
National Archives of Australia, National Archives and Records Administration, The National
Archives) have published papers or web articles on significant properties (also called “significant
characteristics” or “essential characteristics”) in the context of providing a means of measuring
whether a preservation strategy such as migration or emulation is successful, by comparing how
well a target file retains the properties found in the source file. The “Significant Properties
Report” (Wilson, 2007b) provides a useful overview, beginning with a reference to
“Canonicalization: A Fundamental Tool to Facilitate Preservation and Management of Digital
Information” in which Lynch notes, “We want to be able to guarantee that for a given object the
reformatted version is equivalent to the original version with regard to some specific set of object
characteristics” (as quoted in Wilson, 2007b, p.5).
An important shift in the significant properties discussion came with the general
acceptance that digital objects “do not need to remain in a state that is unchanged from their
original state in order for them to be considered authentic” (Wilson, 2007b, p.4). Instead, “A
record is considered essentially complete and uncorrupted if the message meant to communicate
in order to achieve its purpose is unaltered” (The National Archives, 2002, p.8, as quoted in
Wilson, 2007b, p.4). However, an ensuing problem results, because what is considered
“essential” may vary from audience to audience. For example, when looking at medieval
manuscripts, an audience interested in text analysis would consider the text of a document to be
essential, while an audience interested in literary metaphor would insist that the illustrative and
design components as important as the text. Unfortunately, despite a “pressing need” to “develop
a methodology, and begin identifying quantifiable sets of significant properties for specific
classes of digital object[s]” (Wilson, 2007b, p.7), there is no definitive set of significant
properties available.
Although some studies provide examples of significant properties for audio, email, raster
images, and structured text (Grace, 2009), the InSPECT Framework Report reflects a general
move towards developing a methodology or framework whereby “an evaluator operating in a
curatorial institution can determine the properties that they consider to be essential based on their
interpretation of acceptable loss” (Knight, 2009, p.9). To this end, institutions such as the Library
of Congress and the Florida Digital Archives have identified and posted the significant properties
of interest to their institution on their websites.
B. Methods
Essentially, the research project relied selecting a workable number of file viewer
products, formats and files, and comparing each file displayed in the native application with the
same file displayed in a range of file -viewer software products. There were five phases to the
project: product selection; software comprehension; test environment set-up; file formats, file
properties, and files selection; and testing and data collection. With the exception of software
comprehension, all phases were run twice, with adjustments made between the two runs. At the
end of the project, the results from the second run were reviewed to see how well each file
viewer performed, and a determination was made as to whether the file viewer “passed” or
“failed” with regard to each file format.
Product Selection
The main criteria for selecting the file viewer products included affordability, ease-of-
access, and number of format categories covered. These three criteria can be explained as
follows. First, in the previous shared drive migration project, none of the five utility applications
cost more than $100 US per licence, and based on this experience, this amount was used as the
cut-off point for affordability. Second, due to information technology policies at the City of
Surrey, unsupported software cannot be loaded to office computers, and software is only
considered supported after being put through an internal review processes. Lacking a guarantee
that any of the software would be adopted by the co-investigators’ organizational unit, a personal
laptop was used for file viewer loading, meaning that only software compatible with a Windows
environment was loaded, and that higher-level programming abilities were not required for the
install. Third, format categories were loosely considered to be text, data, email, drawing, still
image, and moving image, and any file viewer selected was required to provide services for at
least two categories. Whether or not the file viewer offered integration with the ECM product
was noted, but if the file viewer met the other criteria, it was selected, regardless.
With these criteria in mind, a number of Google searches (e.g., “file viewers,” “universal
viewers,” “best file viewers”) were completed, resulting in a list of possible products. Next,
Download.com and SorceForge.net were used as qualifying resources. Download.com features
software reviews, technology news and software downloads. Their section on universal viewers
helped the co-investigators short list the top 15 products within the cost requirement. Once the
products were short-listed, each product website was reviewed to identify the best fit for the
project, and the final selection was made. SouceForge.net, a site for open-source software
development, was also referenced, but although open-source file viewers were identified, only
one open-source file viewer indicated that two format categories were supported, and an attempt
to download this product was unsuccessful due to programming requirements.
Software Comprehension
Once the file viewers were identified, the product websites were reviewed in an attempt
to find information on how the products work. Since these are proprietary products, little was
available. Next, the co-investigators contacted the software developers directly, using email and
web forums. In every case, the developers were advised that the co-investigators were seeking
information for a research paper on file viewers.
Although there are currently over 40 formats available in the ECM system, only the top
12 formats were selected for testing, with each represented by at least 500 files and up to 18
years’ worth of files (with one exception). In this respect, the file formats selected reflect an
intent to test file viewers, rather than file formats. Perhaps focusing on popular file formats
provided the file viewers products with a better probability of success, but this was considered
acceptable based on the discovery nature of the project. As well, the choice of popular file
formats was dependent on the particular business units whose files were stored in the ECM
system, which currently supports about 10 percent of the networked staff population. The
selection of file formats chosen at another organization might differ (as would a selection made
after more business units are added to the ECM system), but, based on the City of Surrey file
format list, the co-investigators are inclined to believe that many of their popular formats would
testing, only nine files were identified for each of 12 formats for a total of 108 files. Significant
care was taken to ensure that each of the files was an appropriate candidate for testing and
presented properties and characteristics of interest.
Data Collection
Data was collected for using a separate chart for each file format and each file viewer, for
a total of 72 charts (see Tables 3 to 5). These results were tabulated in pass/fail summary charts,
separated into the three time periods (see Tables 6 to 8).
C. Results
the fine detail on this as it is quite a complex subject.” However, by piecing together the
information provided by the software developers, a basic understanding was achieved.
In general, file viewers work by identifying file formats through header information,
magic numbers, or content, and then rendering the content in human-readable form. If the file
format is one that the viewer can render “as is,” the file is displayed in native format. If not, the
file is converted to a second format and then rendered. In order to extend their file format
rendering capabilities, file viewers often consist of a number of viewers bundled together. For
example, the Accessory Software File Viewer contact noted their product uses viewers from
Internet Explorer (for text, html, and Microsoft Object Linking and Embedding or OLE);
LEADTOOLS (for images); and Delphi (for data with open database compliancy or ODBC).
Similarly, UV ViewSoft leverages the Microsoft Internet Explorer engine (for html); a doc-rtf
converter (for text); and Delphi (for data). The Daeja ViewONE contact referred to “third-party
libraries,” and the Stellant respondent noted the use of “outside-in” libraries which convert
“foreign” formats to a generic format which standard viewers can then access. As a result, it
appears that most products rely on file filters and conversion, and are, in the words of the Stellant
respondent “actually rendering a much smaller number of standard formats” than the 100 to 300
file formats commonly listed in product information. During testing, it was interesting to note
that all six products launched Adobe Reader to render .pdf files.
In some cases, the file viewer product is intended for specific format categories—for
example, IrfanView is intended for use with image and audio/video file formats, while
FileStream Turbo Browser extends to six format categories. (As a side note, software products
that render audio/video file formats are referred to as “players” rather than viewers.) The
capabilities noted in product information are summarized in Table 1.
Table 1: File Viewer Products and File Format Capabilities (based on product information)
Moving
Products Text Data Email Drawings Images
Images
Accessory Software File
Yes Yes No No Yes Yes
Viewer
FileStream Turbo Browser Yes Yes Yes Yes Yes Yes
Some file viewers include additional functionality, such as format conversion, file
annotation, file redaction, and integration with electronic record or content management systems.
It is assumed that the software developers producing the file viewers look to appeal to a wider
market, first by extending the number of file formats that their file viewers can render, and,
second, by broadening the marketability of the product by adding related features which users
may be interested in. These features included format conversion, editing, annotation, redaction,
and product integration, and were more commonly available in the category file viewers that
were more costly and likely intended for integration with other software. As shown in Table 2,
few of these features were available in the file viewers selected for testing in this project.
Table 2: File Viewer Products and Additional Features (based on product information)
Format Product
Products Edit Annotation Redaction
Conversion Integration
Accessory Software File
No No No No No
Viewer
IrfanView No No No No Yes
UV ViewSoft No No No No Yes
Based on the product information available, the co-investigators expected that FileStream
Turbo Browser and Quick View Plus to perform the best during the data collection phase of the
project.
identical, sharing the same software load and common access to the shared drive and files.
During testing, differences in screen size and colour space were manually compensated for, and
did not present a barrier to data collection. However, because the personal laptop did not have
access to the shared drives, files had to be copied, resulting in changes to critical metadata (i.e.
Owner, Date Created, Location). This would not have been the case if both computers were
networked workstations.
Mitigating the problem of changed metadata by confirming its availability and values
reinforced the co-investigators’ belief that metadata properties are different from file
characteristics. From their point-of-view, a file viewer does not need to be able to access
properties that can be better assessed with a more suitable tool such as an operating system or a
file manager utility application.
File Formats, File Properties and Characteristics, and Files Selection Results
The twelve file formats chosen for the second run of testing included: text (.doc, .pdf,
.ppt), data (.xls), email (.msg), web (.htm), drawing (.dwg, .vsd), image (.jpg, .tif), and moving
images (.mov, .avi). All of these formats met the requirement of at least 500 files and at least
three files from each of the three time blocks (with the exception of the .avi format, where only
files from the last time block were found).
During preliminary testing, a discovery was made that four out of the six viewers could
not render Microsoft files in “.x” file formats (i.e., .docx, .pptx, .xlsx), designed to meet the
Office Open XML standard. Additionally, the file viewers that could open .htm files rendered the
files as text representations with style tags, without graphic representation. The reason for the
“xml” gap in the file viewers is not known. Perhaps the developers of these products do not
consider xml file formats problematic, assuming that these files will be viewed using a web
browser or editor. Or perhaps the xml file formats are too new, and the development work is not
complete. At any rate, the .x file formats were removed from the test sample.
Regarding properties (determined without opening a file), there were few unexpected
results. Word Count was not available for view for .pdf files in either Windows or file manager.
The same issue appeared for Length, Width, Height, Pixel Aspect Ratio, and Frame Rate for
.mov files.
Regarding characteristics (determined by opening a file), there were a number of general
observations. First, where a file viewer could render a file on screen, the file could be
successfully sent to Print. Second, the web file format (.htm) and image file formats (.jpg, .tiff)
were rendered by all six file viewers, although it should be noted that graphic representation of
the web file format (.htm) was not a mandatory requirement. Third, some characteristics, such as
slide presentation and animation (.ppt) or formulas, macros, and links (.xls) were not represented
by any of the file viewers. Because these characteristics appeared to present a common problem
to all viewers, they were treated as non-mandatory. Perhaps these characteristics are proprietary
features of the native applications, and require execution rather than simple rendering. Fourth, in
most cases, if a file viewer could render a file format, it could open older versions of the file
format. Exceptions included GetData Explorer View (for .ppt in 2000-2005 and 1994-1999 and
for .xls in 1994-1999) and UV ViewSoft Viewer (for .doc in 1994-1999).
During testing, four .docx files were identified and replaced from the test pool after five
file viewers failed to render the files. An additional two files were identified as corrupt as they
could not be opened in the native application. These files were removed and replaced.
Data Results
Files were selected on the basis that mandatory properties were present and viewable in
the operating environment and/or the file manager utility application and the native application.
To ensure that good sample files were selected for the test pool, the data was extracted and
collected in summary tables for each file format (see Table 3). In some files, the co-investigators
found that mandatory properties were missing or overwritten. In these cases, the files were
discarded. A total of 72 properties tables were created (i.e. six file viewers x 12 formats).
Table 3: Operating System Results for .doc Properties (mandatory and optional)
File 3 DCT Old Pre 7 4 Documents Cognos 1.doc SURREY\IAM 2000-01-17 07:10 363
2006 to 2011
DCT IP3 Creator Preserver Responsibilities
File 1 SURREY\LE2 2009-05-22 13:03 3188
V 03 0.doc
SOW Storage Solution Facilities Plans 2008
File 2 SURREY\LE2 2008-08-26 07:27 935
08 25 v01 0.doc
File 3 DCT Master List 2011.doc SURREY\EAG 2010-12-02 11:00 3051
Once the file properties were checked, each format was tested on each file viewer by
opening each file in the native application on the networked workstation, and then opening a
copy of the file on the personal laptop. The characteristics of the files were compared, and the
results were recorded in 72 separate tables (see Table 4).
Table 4: File Viewer Showing Pass Results for Characteristics for .doc File Format
Table 5: File Viewer Showing Fail Results for Characteristics for .doc File Format
Once the data for characteristics were collected, the pass/fails for each file viewer and file
format were compiled into three charts, showing the performance of the file viewer on newer
files dated from 2006 to 2011, somewhat older files from 2000 to 2005, and older files from
1994 to 1999 (see Tables 6-8).
There were no file viewers that could successfully render all 12 file formats. Two file
viewers were able to open 10 out of 12 formats: FileStream Turbo Browser and Quick View
Plus. Turbo Browser was unable to open .doc and .vsd files, Quick View was unable to open
.mov or .avi files. Of these two products, Quick View performed more closely to product claims
as interpreted by the co-investigators (see Table 1).
File Viewer DOC PDF PPT XLS MSG HTM DWG VSD JPG TIFF MOV AVI
Accessory Software
PASS PASS FAIL PASS FAIL PASS FAIL FAIL PASS PASS FAIL PASS
File Viewer
FileStream Turbo
FAIL PASS PASS PASS PASS PASS PASS FAIL PASS PASS PASS PASS
Browser
GetData Explorer View FAIL PASS PASS PASS FAIL PASS FAIL FAIL PASS PASS FAIL FAIL
Irfan View FAIL FAIL FAIL FAIL FAIL PASS FAIL FAIL PASS PASS FAIL PASS
Quick View Plus PASS PASS PASS PASS PASS PASS PASS PASS PASS PASS FAIL FAIL
UV Viewsoft Viewer PASS PASS FAIL PASS FAIL PASS FAIL FAIL PASS PASS PASS PASS
File Viewer DOC PDF PPT XLS MSG HTM DWG VSD JPG TIFF MOV AVI
Accessory Software
PASS PASS FAIL PASS FAIL PASS FAIL FAIL PASS PASS FAIL N/A
File Viewer
FileStream Turbo
FAIL PASS PASS PASS PASS PASS PASS FAIL PASS PASS PASS N/A
Browser
GetData Explorer View FAIL PASS FAIL PASS FAIL PASS FAIL FAIL PASS PASS FAIL N/A
Irfan View FAIL FAIL FAIL FAIL FAIL PASS FAIL FAIL PASS PASS FAIL N/A
Quick View Plus PASS PASS PASS PASS PASS PASS PASS PASS PASS PASS FAIL N/A
UV Viewsoft Viewer PASS PASS FAIL PASS FAIL PASS FAIL FAIL PASS PASS PASS N/A
File Viewer DOC PDF PPT XLS MSG HTM DWG VSD JPG TIFF MOV AVI
Accessory Software
PASS PASS FAIL PASS FAIL PASS FAIL FAIL PASS PASS FAIL N/A
File Viewer
FileStream Turbo
FAIL PASS PASS PASS PASS PASS PASS FAIL PASS PASS PASS N/A
Browser
GetData Explorer View FAIL PASS FAIL FAIL FAIL PASS FAIL FAIL PASS PASS FAIL N/A
Irfan View FAIL FAIL FAIL FAIL FAIL PASS FAIL FAIL PASS PASS FAIL N/A
Quick View Plus PASS PASS PASS PASS PASS PASS PASS PASS PASS PASS FAIL N/A
UV Viewsoft Viewer FAIL PASS FAIL PASS FAIL PASS FAIL FAIL PASS PASS PASS N/A
D. Conclusion
The project provided relatively straight-forward results for the first three research
questions. First, file viewers were found to work by rendering a file in its native file format, or
by converting a file to a second format that could be rendered. Second, file-viewer software
products appear to fall into two categories: lower-cost, stand-alone products; and higher-cost
products capable of integration with electronic document, record, or content management
systems. In general, the six lower-cost, stand-alone products tested performed as per the product
claims. Differences between product claims and performance may depend on the metrics chosen
be viewed from within a file is, in the minds of the co-investigators, irrelevant. If the properties
exist, are unchanged, and can be extracted using the operating system or a file manager, their
maintenance is likely a separate issue than whether or not data characteristics are evident in a
migration copy. Characteristics need to be rendered to be observed, whether rendered on the fly
by a file viewer or created through conversion. Based on this understanding, the co-investigators
suggest that a file-viewer approach has a minor advantage over a migration-conversion approach
as the original properties are evident as they are embedded in the single file. In migration
conversion, the relationship between both the files must be created and maintained, along with
both sets of properties.
Returning to characteristics, the file viewers displayed similar issues around data loss as
experienced in migration conversion. In this respect, it is interesting to note that where the
original developer’s rendering software was available, all six file viewer developers adopted it as
the rendering “engine” for that format. During testing, the co-investigators found that where a
.pdf file was launched using a file viewer, the file opened in Adobe Acrobat Reader. One
wonders, if all software developers made their rendering engines available (as Adobe has),
whether file-viewer developers might not be able to create products closer to a “universal”
viewer, capable of rendering all (or at least more) formats. Additionally, might not some of the
characteristics that render (or play) within the native application and that are not apparent in file
viewer or migration conversions be made available from within the rendering engines to provide
a fuller data range for these objects? After all, the Adobe Acrobat business model of providing
the “reader” for free and charging for the “writer” proved successful, even after the .pdf
specification was made available. Although some type of planetary alignment might be required
to achieve this effect, it is not impossible, given the number of archives, records management,
and library specialist now involved in standards boards and other organizations that facilitate
access to software development companies.
In this respect, it may be of value for the archives and records management community to
continue to clarify their requirements in terms of properties and characteristics. Properties, in the
sense of file property metadata, are clearly conveyed through standards, data dictionaries, and
many other forums, but additional clarification around characteristics would be useful. Here, the
co-investigators observe three categories of characteristics: structure-related (e.g., cells, line
breaks, page breaks, tables, and bullets); appearance-related (e.g., font size and colour, images,
and diagrams), and behaviour- related (e.g., formulas, macros, and slide presentation and
animations). Similar observations regarding categories of significant properties were noted in the
InSPECT Significant Properties Report (Wilson, 2007b), including content, context, appearance,
structure, and behaviour, but no separation was made between properties and characteristics.
In closing, the co-investigators recognize the migration-conversion approach as the
primary digital preservation strategy in place in most archive institutions today. This strategy
provides important risk insurance for digital objects, and especially those in danger of immediate
obsolescence. For some organizations, the risk of not having electronic information available in
an accessible format largely outweighs the total costs of file migration. However, other
organizations may find that the migration-conversion strategy is not always a viable option, and
they still may need to provide some means of long-term preservation of digital objects. The file
viewer approach was reviewed by the co-investigators in the face of specific problems in
introducing migration conversion to their environment. Additionally, the migration-conversion
strategy is not perfect, as data characteristics are often lost. With many institutions maintaining
the native files as their only record or in addition to a preservation copy, an opportunity exists to
pursue other, complementary strategies. For these two reasons, the co-investigators suggest that
file viewers provide an opportunity to leverage native files in a less resource-intensive manner.
The co-investigators note also the extensive body of work in progress beyond the field of
archives and records management, and the need to collaborate with other fields, including
software development in the pursuit of the file viewer and other digital preservation strategies.
E. Acknowledgements
The co-investigators would like to acknowledge the work of the InterPARES 3 Graduate
Research Assistants, Jen Busch and Sergey Kovynev, who participated in launching the General
Study on File Viewers, and the contributions of Dr. Luciana Duranti, InterPARES Project
Director.
F. References
“Accessory Software File Viewer 9,” available at http://www.accessoryware.com/FileView.htm
(accessed November 2011 - February 2012).
Amirani M., Toorani M. and Shirazi A. (2008), “A New Approach to Content-based File
Type Detection,” in Proceedings of the 13th IEEE Symposium on Computers and
Communications (ISCC’08), July 2008, pp. 700-705.
Enns, L. and Badesha, G. (2011), Shared Drive Migration Toolkit, City of Surrey, available at
http://www.interpares.org/ip3/ip3_cs14_report.cfm (accessed September 2011 - February
2012).
The Florida Centre for Library Automation, (2009), “Florida Digital Archives,” available at
http://fclaweb.fcla.edu/FDA (accessed November 2011 - February 2012).
Grace, S., Knight, G. and Montague, L., (2009), “Investigating the Significant Properties of
Electronic Content over time: Final Report,” Centre for e-Research, King’s College London, 21
December 2009.
Knight, G. (2009), “Investigating the Significant Properties of Electronic Content over time:
Framework Report,” Centre for e-Research, King’s College London, 13 October 2009.
Library of Congress, (2010) “Sustainability of Digital Formats: Planning for Library of Congress
Collections,” available at http://www.digitalpreservation.gov/formats/ (accessed November 2011
- February 2012).
McDaniel, M. and Heydari, M. H., (2002), “Content Based File Type Detection Algorithms,”
available at http://dl.acm.org/citation.cfm?id=821828 (accessed November 2011 - February
2012).
McHenry, K., Kooper, R. and Bajcsy, P. (2009), “Towards a Universal, Quantifiable, and
Scalable File Format Converter,” available at
http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=5380873 (accessed November 2011
- February 2012).
Mellor, P., Wheatly, P. and Sergeant, D., (2002), “Migration on Request, A Practical
Technique for Preservation,” in Research and Advanced Technology for Digital Libraries, 6th
European Conference, ECDL 2002, Proceedings, Rome, Italy, 16-18 September 2002, pp.
516-526.
Rogers, C., Malmas, S., and Enns, L., (2011), “Case Study Final Report: CS 14 City of
Surrey Policies, Guidelines and Procedures for a Drive Migration Project as part of an
Enterprise Content Management Program,” available at
http://interpares.org/ip3/display_file.cfm?doc=ip3_canada_cs14_final_report.pdf.
Roseman, M. and Greenberg, S., (1995), “Building Real-time Groupware with GroupKit , A
Groupware Toolkit,” ACM Transactions on Computer-Human Interaction (TOCHI), Volume
3, Issue 1, pp. 1-30.
Skiljan, I., (2005), “Irfan View,” available at http://www.irfanview.ca/ (accessed November 2011
- February 2012).
Wilson, A. (2007b), “InSPECT: Significant Properties Report,” Arts and Humanities Data
Service, 10 April 2007.