The Archived Website and Website Philology A New Type of Historical Document?

Abstract

Website history can be considered an emerging discipline at the intersection between media history and Internet history. In this discipline, the individual website is regarded as the unifying entity of the historical analysis rather than the Internet or the Web. Writing the history of a website involves using many sources and methods similar to those used in writing the history of any other media type. But one document type requires special attention: the archived website. This is so because the problems involved in finding, col - lecting and preserving the website are different from those characterizing the archiving of other types of traces of human activity, including other media types. The primary problem is that the actual act of finding, collecting and preserving changes the website that was on the live web in a number of ways, thus creating a unique version of it and not simply a copy. The present article sets out, first, to discuss to what extent the archived website can be considered a new type of historical document and how its characteristics affect the task of the website historian who must later use it; second, the article discusses and attempts to formulate some methodological principles, rules and recommendations for a future critical textual philology of the website.

Full text

155
Nordicom Review 29 (2008) 2, pp. 155-175
The Archived Website
and Website Philology
A New Type of Historical Document?
Niels Brügger
Abstract
Website history can be considered an emerging discipline at the intersection between
media history and Internet history. In this discipline, the individual website is regarded as
the unifying entity of the historical analysis rather than the Internet or the Web. Writing
the history of a website involves using many sources and methods similar to those used
in writing the history of any other media type. But one document type requires special
attention: the archived website. This is so because the problems involved in nding, col-
lecting and preserving the website are different from those characterizing the archiving of
other types of traces of human activity, including other media types. The primary problem
is that the actual act of nding, collecting and preserving changes the website that was on
the live web in a number of ways, thus creating a unique version of it and not simply a
copy. The present article sets out, rst, to discuss to what extent the archived website can
be considered a new type of historical document and how its characteristics affect the task
of the website historian who must later use it; second, the article discusses and attempts to
formulate some methodological principles, rules and recommendations for a future critical
textual philology of the website.
Keywords: internet, web, website, media history, archiving, archive, philology
Introduction
Website history can be considered an emerging discipline at the intersection between
media history and Internet history. In this discipline, the individual website is regarded
as the unifying entity of the historical analysis rather than the Internet or the Web.
Website history is a natural part of existing media history; the website can partly be
seen as the latest acquisition in the history of co-evolving types of media, from print
and analogue media to a whole range of digital media, and website history can partly
focus on the same analytical domains as existing media histories have done throughout
the years, such as institutional analysis, reception studies, studies of media culture and
text analysis in a broad sense (analysis of individual programmes, genres, ow, range
of programmes, etc.)1 However, to some extent the traditional conceptual frameworks
need to be re-evaluated.2
Website history is also an integral part of both the Internet and Web history. Internet
history can be considered the broadest category – covering the history of, for instance,
email, usenet, mailing lists, IRC etc. – while web history is a sub-discipline within In-

156
Niels Brügger
ternet history dealing with the history of the Web as such, the bloggosphere and so on.3
Website history, for its part, is a sub-discipline of Web history, and its subjects coincide
at least in part with those of both Internet and Web history, as the history of a given
website is closely related to the history of both the Internet and the Web. Nevertheless,
the history of a website cannot be understood exhaustively solely by using concepts and
insights from Internet or Web history (cf. also Brügger 2007a).4
Writing the history of a website involves using many sources and methods similar
to those used in writing the history of any other media type. For instance, the research
project ‘The History of www.dr.dk, 1996-2006’ (cf. below) will be based on such sources
as memos, minutes of meetings, reports, policy documents, correspondences, organiza-
tional charts, job descriptions, dummies, retrospective research interviews, biographies,
diaries, memorabilia, legal texts and statistics.
But one document type requires special attention: the archived website. This is so
because the problems involved in nding, collecting and preserving the website are dif-
ferent from those characterizing the archiving of other types of traces of human activity,
including other media types.
The primary problem is that the actual act of nding, collecting and preserving chan-
ges the website that was on the live web in a number of ways, thus creating a unique
version of it and not simply a copy. Therefore, it could be argued that the archived web-
site constitutes a type of historical document, which in many ways differs signicantly
from other well-known document types.
The present article sets out, rst, to discuss to what extent the archived website can
be considered a new type of historical document and how its characteristics affect the
task of the website historian who must later use it; second, the article discusses and
attempts to formulate some methodological principles, rules and recommendations
for a future critical textual philology of the website. Satisfactory answers to these two
clusters of methodological questions are necessary stepping-stones on the path towards
writing website history.5
Finding, Collecting and Preserving Websites
If we maintain that the act of nding, collecting and preserving changes the website that
was actually on the live web, it becomes relevant to identify the reasons for this asser-
tion.6 In overall terms, one can distinguish between six ways of integrating web material
in a web archive, and these six ways may be split into two main groups depending on
whether the material is archived from the Net or delivered from the producer (some of
the sub-groups may overlap).
1. Archiving from the Net:
a. Harvesting, capturing or lming
b. Harvesting, capturing or lming ‘ghost websites’
c. Harvesting, capturing or lming websites archived on the Net by the producer
2. Delivery:
a. Delivery of non-archived material from the producer
b. Delivery of archived material from the producer
c. Delivery from other archives7

157
The Archived WeBsiTe ANd WeBsiTe Philology
A Subjective Re-construction
Any kind of website archiving, whether done by means of harvesting, capturing and
lming or delivery, is a re-construction based on the bits and pieces that stem from either
the Web or the producer of the website. And this re-construction involves a number of
subjective choices and unpredictable coincidences with regard to software, strategy and
purpose, integration in an archive and so on (cf. Brügger 2005: 15-19, 30-31, 61-62; cf.
also Schneider & Foot 2004: 115; Masanès 2006: 17-18, 76).
For instance, it must be decided whether one intends to harvest, capture or lm the
website, or employ a combination of these three methods, and if so, which combination?
And does one intend to archive the entire website? How many levels should the archiving
cover? Are photos, sound and moving images to be included? Is material to be collected
from other servers? One must also decide the size of what could be called ‘the archival
element’ (cf. Brügger 2005: 40) and the archiving direction; in other words, how do
we intend to ‘slice up’ the website? By using a variety of start URLs on more than one
level? And in what order are the archivings from the start URLs to begin? And, nally,
a number of choices must often be made with respect to the subsequent montage of the
archived material.
In this way, the archiving of a website stands apart from the archiving of other types
of objects, be they physical objects, including types of media such as the printed media,
photographs and lm, or electronic ‘objects’, from radio and TV broadcasts to cellular
telephone services. Regardless of who stacks yesterday’s newspapers at the library, they
look the same, just as the TV programme looks the same regardless of who puts in the
tape and presses the record button. The subjective element almost exclusively lies in the
act of selecting rather than in the archiving process itself, which by and large consists
of taking the objects out of circulation and preserving them as they were, unchanged
(cf. Brügger 2005: 15-19).8
In contrast, archiving a website is an active process from the very beginning, where
to some extent the archived website does not exist prior to the act of archiving; it is only
created through the archiving process, on the basis of ‘raw material’ from the Internet or
the producer (elsewhere I have used the expression ‘a document of the Internet’ to refer
to the archived (harvested) website, cf. Brügger 2005: 30). Two overall and interrelated
consequences follow from this. First, the processes of research – for instance, the writing
of website history – and archiving are closely connected, more than we know to be the
case for other types of media. Second, archiving a website should be accompanied by
deliberations as to method: Why and how has the archived version been created? (cf.
Brügger 2005: 31-32, and Schneider & Foot 2004: 115).
The fact that the archived website is a subjective re-construction constitutes a fun-
damental condition of all of the six distinct ways of integrating material in a website
archive. But besides this, each of them is characterized by specic problems.
Archiving from the Net
The problems related to website harvesting have set the agenda for discussions of web
archiving for the past ten years, and by now many of the problems have been detailed
and to a certain degree solved (for reasons of space, only the problems related to harves-
ting are in focus in the following pages; problems related to capturing and lming are
discussed in Brügger 2005: 47-60).9 However, the dynamic character of the Internet will
constantly be a source of uncertainty in relation to harvesting, above all, the dynamic

158
Niels Brügger
that I have called ‘the dynamic of updating’ (Brügger 2005: 21-27; cf. also Schneider &
Foot 2004: 115). The dynamic of updating refers to the fact that the content of a website
may change during the process of archiving, and we do not know if, where, and when
the updates occur, which has the following two consequences:
The first and rather obvious consequence is that we cannot be sure that we have
everything in our archive. We will always have lost something in the asynchronous
relationship between updating and archiving. The second consequence is less
obvious, but no less serious. Not only do we lose something that was there, we
are also in danger of getting something that in a way was never there – something
that is different from what was really there. My archived version of a newspaper’s
website can be a combination of elements from two (or more) versions that were
there at different times – but they were never there at the same time as they might
now be in my archive. We thus face the following paradox: on the one hand, the
archive is not exactly as the website really was in the past (we have lost something),
but on the other, the archive may be exactly as the Internet never was in the past
(we get something different). (Brügger 2005: 23).10
A variant of this problem is the fact that the greater the time span between two harvests
of a website, the more difcult it is to date all elements on the website to a precise
point in time. Thus, we cannot know to what extent the harvesting time is identical to
the time of publication; in other words, we do not know how long the website, in all its
dimensions, has been the way it is when the harvesting takes place. What is harvested
is both a point in time (the time of harvesting) and a period of time (the period up to
the time of harvesting). This characterizes all kinds of cross-sectional harvests made at
considerable intervals.
The fact that evidence of human activities in the past cannot be dated to a precise
point in time, or that the evidence might have been modied within a certain period of
time, is in itself not new. Very often it is only possible to date utensils, documents and
the like to a given period of time – an antique Greek vase was used between 200 and
100 BC; a document was written between the 8th and 9th centuries – but within the
limits of this period of time, the object is identical to itself and is characterized by a
one-dimensional temporal logic: It is the same vase or the same document; we are just
unable to say precisely when it was created or used within the period. And even if the
object has been subject to changes – the vase may have been re-decorated, the medieval
hand-written manuscript overwritten, or pages added – the changes will almost always
have left some datable traces, allowing us to date them within the period, based on a
one-dimensional temporal logic with a clear ‘before’ and ‘after’.
The situation is different for a website harvested at considerable intervals of time or
for the rst time. In this case, we have an end date (the time of harvesting) and perhaps a
kind of start date (an earlier version), but within this period of time a continuous process
of publication may have taken place, and it is very difcult to say if, where, and when
this occurred, as in most cases publication does not leave any datable traces. Therefore,
such an archived website may not be identical to itself within the period that can be
delimited, and it is based on a multi-dimensional temporal logic: some elements on the
archived website are from one point in time, other elements from another point in time.
Imagine a preserved newspaper composed of bits and pieces from several newspapers
from the period prior to the copy in question, but without any temporal indication on
the fragments. Concerning the website, we might even encounter a temporal logic that

159
The Archived WeBsiTe ANd WeBsiTe Philology
not only extends back into the past, but also forward into the present or the future – for
instance, when an older website contains commands that are set to get the current news
headlines, weather forecast or the like from a web server, and will therefore get the news
from the date of harvesting, and later on perhaps even present the news from the date
the archive is consulted – like a year-old newspaper in the library suddenly showing
today’s weather forecast.
If we look at the second way of integrating web material in an archive, the harvesting
of ghost websites, the problem is not the harvesting itself, or rather: the problems are not
greater than or different from those encountered in harvesting in general. The problem is
partly nding the material, partly dating it. First, we have to become aware of the exis-
tence of the material – here we often have to rely on other types of sources such as press
releases, other media and other sub-sites. In this task, the producer will not always be
much help, because systematic policies for discarding old material are rare. Second, we
have to date a website that is harvested for the rst time, and, as shown above, it may be
difcult to determine how long it has looked the way it does at the time of harvesting.
The problems related to the third method, harvesting websites archived on the Net by
the producer, resemble those characterizing the ghost websites; however, the problem of
nding the material will probably be less signicant, as it has been saved intentionally,
which is why there will normally be links to it, just as it will often be easy to identify
as archived material if it is marked with watermarks or the like.
Delivery
While harvesting has been setting the agenda in discussions of web archiving for some
years, the problems that might be related to the delivery of old web material are unk-
nown quantities, at least to my knowledge.11 Hence, what is written about delivery in the
present article is not based upon thorough discussions and experience; it should rather
be considered a provisional catalogue of the possible problems.
In general, the fact that the archive is not in control of the formats of delivery (soft-
as well as hardware) will constitute a problem for all three methods of delivery. First,
reading the material can be problematic: File types may have gone out of use or it may
be impossible to establish an adequate computer environment.12 Second, integrating
delivered material into existing archival structures and procedures can be problematic:
must GIF, PDF or QuickTime les be converted to HTML or ARC les? Can an archive
from 2007 integrate material from 1996 without automatically dating it to 2007? How
can the material be made searchable if it is in a different format than the les of the
archive and if it does not provide any kind of text (e.g., no URL, HTML); and must
searchable data then be provided manually by the archive? How can material already in
the archive be made to interact with delivered web material of the same website from
the same date (e.g., individual les, sub-sites)? And so on and so forth.13
The problems related to the delivery of web material are supposedly not yet very
great, which is probably a consequence of the fact that the number of websites archi-
ved by individual scholars (micro or on-demand archiving) is still rather limited. But
assuming that website history – and website studies in general – will become more
widespread, and that as a result the number of archives will grow, we may witness an
increased demand for the mutual exchange of archived web material among scholars
(micro to micro), on the one hand, and, on the other, retrospective collecting initiatives
launched by the great (inter)national archiving institutions with the aim of supplementing

160
Niels Brügger
and completing the cultural heritage (micro to macro).14 Hence, we have good reason to
put the problems related to the delivery of web material on the agenda.
If an archive is confronted with the rst method of delivery – delivery of non-archived
material from the producer – it will probably nd the complex, fragmented, varied or
unpredictable character of the material problematic. First, it may be difcult to re-create
a meaningful unity – a web page, a website – out of a pile of heterogeneous bits and
pieces.15 Second, dating the material may constitute a problem. In contrast to harvesting,
the problems are even more difcult in this respect because, unlike the time of harves-
ting, one cannot even be sure of having one xed point in time. With harvesting, the
problem is not determining a period of time but rather the possible lack of simultaneity
and self-identity within the period, but the problem with the delivered material discussed
here is that it can be difcult to ‘insert’ a time period in the fragmented continuum of,
for instance, a collection of graphic les, a CMS database or the like. Thus, the problem
is not (only) dating the material on a time line, but rather re-creating something that
can be related to a time line in a meaningful way, as we cannot be sure that there is any
consistent division in either the points or periods of time inscribed in the material.
Different kinds of problems can be expected to arise from the other method of de-
livery: delivery of archived material from the producer. It will probably have a more
systematic character, but this may also constitute a complicating factor if, for instance, an
entire computer environment has to be established in order to restore a backup version.
It will probably be impossible to integrate such an environment in an existing archive,
which is why the archive might have to harvest the restored version. The result will be a
kind of ‘double archiving’ where the website is both re-created and harvested, resulting
in two archiving phases, each characterized by specic subjective choices. Besides,
documentation may be lacking, which can make it difcult to use the material.
The last method of delivery, from other archives, can entail all the problems men-
tioned above, but they may be aggravated in connection with collections from micro
archives, because the variations in, for example, archiving purposes and the use of ar-
chiving software are probably greater among ‘amateurs’ than professionals.16 Besides,
the integration of corpora – material that has been archived as a whole, often to be used
for a specic purpose in a research project – can constitute a special problem. In overall
terms, a corpus resemble the strategy of event archiving, and in relation to this kind of
material, it is important that the archived websites be marked as part of a corpus so that
future users know why the material looks the way it does, as it was created with a specic
purpose in mind; any supplementary documentation should also be supplied (e.g., the
research questions, plans for the archiving process, the archiving log).
Besides all the problems emanating from both archiving the Net and the delivery of
old web material, an overall problem for the archive should be mentioned: that of making
the many different types of material in each of the six methods interact in a useful and
meaningful way in the same archive, from both a technical and a conceptual point of
view (cf. the question of the montage of smaller web elements into a meaningful unity
is discussed in Brügger 2005: 39-62).
What Can We Expect to Find in a Web Archive?
As has been argued above, the well-known processes of nding, collecting and preser-
ving appear to be different when the object in question is a website, both on a general
level and when we look more closely at the six specic ways of integrating web material

161
The Archived WeBsiTe ANd WeBsiTe Philology
in a web archive. However, if one wishes to write website history, this is the way things
are, and as website historians, we cannot afford to leave the many bits and pieces of
web material without trying to make them useful as sources by integrating them into an
archive, one way or another.
If the considerations above are taken as a starting point, then what can the historian
who intends to write website history expect to nd in a website archive? He or she can
expect two things. First, that if several archived copies of a given website exist from
the same date, they are very likely to be different from one another (to various degrees).
This is because of the elements of subjectivity, creation and coincidence, which generally
characterize the process of archiving or integrating a website into a website archive,
just as the problems related to each of the six specic ways of integrating material
into an archive also play a role. Second, the website historian cannot expect to nd an
original in the form of the website as it actually looked on the Internet at a given time,
neither in the sense of nding an original among the different versions nor in the sense
of reconstructing an original on the basis of the different versions. In both cases, this
is because a version in the archive may either be lacking something (an ‘incomplete’
original), or it may consist of something that has never been on the Internet at the same
time (a kind of ‘incorrect’ original due to a possible asynchrony between the website
on the Net and the archived version) – and we have great difculty in determining with
certainty if and to what extent a given version is either ‘incomplete’ or ‘incorrect’. In
addition, some archiving software writes in the archived les (e.g., HTTrack), while
other software does not (e.g., Heritrix).
Test of Versions in Existing Web Archives
If this is what can be expected, the question then arises of whether this is really what
one nds when confronted with actual existing archives? In the spring of 2007, the
research project ‘The History of www.dr.dk, 1996-2006’ was started.17 In the context of
the project, a pilot project entitled “Method study of the integration of Internet material
directly received from producers” was being carried out, and part of this pilot project
was a test examining the appearance of a website that had been archived on the same date
(and, if possible, at the same time) in different archives, where it had been integrated on
the basis of harvesting, capturing, lming or delivery from the producer.18 Even if only
four versions from one date have been tested thus far, the tendency is clear: with one
exception, the four versions showed great differences in all respects.
A Critical Textual Philology of the Website
Because we can expect, on the one hand, that if several archived copies of a given
website exist from the same date, they are very likely to be different from one another,
and on the other hand, that an original in the form of the website as it actually looked
on the Internet at a given time cannot be found, knowledge of an original website to
which the different versions can be compared is still wanting, and we must therefore
make do with relative comparisons based on an analysis of the differences and simila-
rities between the existing copies. Therefore, if we intend to say something about how
a given website looked at a given time on the Web – and if we want to do so on the
basis of one or more archived websites – we cannot determine this with certainty, but
only with various degrees of probability. Thus, professing to say what is true or false

162
Niels Brügger
is not an option, instead we must make do with maybe/maybe not. In this task, we are
very close to textual criticism of the philology of manuscripts, in the sense of both
manuscript books and draft manuscripts. What might begin to emerge could be called
a critical textual philology of the website.
Today, the number of website archives is rather limited, and as a consequence the
number of different versions of the same website is also limited. But the problems rela-
ted to a critical textual philology of the website are still a matter of principle, and their
practical relevance is expected to grow as more archives are established, professional
as well as non-professional. First, an increased number of archives will also increase
the number of versions. Second, one can expect two opposing developments to take
place: On the one hand, we can expect more – and more skilled – professional Internet
archives, which means better qualities of versions, but on the other hand, we can also
expect more non-professional archives to be made by website scholars, which means
more versions made by amateurs as well as more versions made for special purposes
(e.g., corpora made for a specic research project).
What follows is a brief outline of some of the methodological principles, rules, and
concepts that a critical textual philology of the website could use to consistently say
something about the probability of how a given website looked at a given time on the
Web.19 The considerations are based upon my work with archived websites from various
archives – professional as well as non-professional archives – during the past six years,
and their applicability and usefulness are being tested as part of the test mentioned
above.
A number of the fundamental concepts and questions in ‘classical’ textual philology
are relevant to a philology of the website, but they have to be brought into line with
the archived website as a type of document based on a specic media materiality that
is different from papyrus, parchment, paper, and printed books.20 However, it should
be stressed that the present article has no intention of giving a thorough presentation
and discussion of textual philology in general, its epistemological history, different
methods, schools, etc. The following lines are meant to be nothing more than a brief
outline of some general lines of demarcation between ‘classical’ textual philology and
website philology.21
From Manuscripts to Archived Websites
The objective of textual philology is basically to examine the differences and similarities
between variants of textual evidence that resemble one another – fragments as well as
entire works – with a view to either nding the most correct variant(s) or reconstructing
a lost original, and to comment on these versions so as to give a critical account of their
nding or creation and to make them understandable for future readers.
Within the textual philology of ‘old’ media, the variants being compared can vary
in two respects. First, as regards the media type, one can compare variants written
on papyrus, parchment or paper to each other (e.g., manuscript books or manuscript
drafts), variants written on papyrus, parchment or paper to the printed version, and -
nally one can compare different printed variants. Second, with respect to the process of
(re)creation of the variants, we can compare variants that are copies of one another (e.g.,
different copies of a specic manuscript book), one draft to another, a specic copy or a
draft to the printed variant, and a printed variant to another printed variant (cf. Fig. 1).

163
The Archived WeBsiTe ANd WeBsiTe Philology
Figure 1. Comparisons of Variants with Respect to Media Type and the Process of
(Re)creation
Media type
manuscript to manuscript to printed book
Process of (re)creation manuscript printed book to printed book
copy to copy +
draft to draft +
copy/draft to printed version +
printed version to printed version +
Figure 1 illustrates that as we move horizontally from manuscript book (written on
papyrus, paper, or parchment) to the printed book, and vertically from various copies
to the printed version, we move at the same time away from the endless copying and
changing towards the stable, authoritative printed version of a book, which constitutes
an original in relation to which all earlier copies can be evaluated, and to which all later
copies are identical.22 In this perspective, a textual philology of the website is close to
the comparison of a copy to a copy, without any authoritative original at hand, as it is
known from manuscript books or manuscript drafts.
The manuscript scholar who today is faced with a number of different ancient
manuscript books or drafts knows for certain that they are identical to the ma-
nuscripts that actually were created and circulated in the past (given, of course,
that they are not forgeries,). He or she ‘simply’ has to examine: a) the possible
differences and similarities between them in order to determine if (and how) one
or more of them constitute a source text or an earlier variant of a draft; b) the
provenance and afliation of the different variants backwards in time, because
the question of a possible source text is related to variants that succeed one
another in the past (cf. Fig. 2).
Figure 2. The Examination of Manuscripts
Note: The arrows indicate the direction of examination.
The website philologist, for his/her part, is subject to quite another media materiality,
which is why he/she has to deal with other problems. His/her task is different from that
ORIGINAL
COPY
Original
COPY
Original
COPY
Original
Copy
Copy

164
Niels Brügger
of the ‘classical’ textual philologist in at least the following seven distinct ways (cf.
Fig. 3 for the rst three points).
1) Because he/she has reason to believe that none of the versions of the archived web-
sites are identical to what was actually on the web in the past, he/she cannot ‘just’
examine the eventual differences and similarities in order to determine if one/more
of them constitute a source text.
2) In contrast to the ‘classical’ textual philologist who examines versions succeeding
in time, the website philologist deals with archived websites that are more likely to
be from almost the same point in time (day, hour, etc.), which is why he/she has to
trace differences and similarities in simultaneity instead of tracing provenance and
afliation backwards in time.
3) Insofar as the website philologist can be said to examine things backwards in time,
he/she only takes one step back, whereas the ‘classical’ philologist in principle can
extend his examination several steps back into the past.
Figure 3. The Examination of Archived Websites
ORIGINAL
COPY
COPY
COPY
Note: The arrows indicate the direction of examination.
4) Due to the materiality of digital writing, the object of the textual philologist of the
website – the text – can be examined on several levels. On the one hand, there is the
immediately perceptible level, where we see/hear the signifying units directly on the
screen or in the speakers. On the other hand, there is the variety of underlying textual
levels that are not immediately perceptible, but nevertheless make possible what
we see/hear: above all, the source code (HTML, XML, etc.), but also the different
layers of the Internet (the TCP/IP model, the OSI model or the like), layers that are
all texts written in a digital alphabet with only two letters, 0 and 1 (cf. Finnemann
1999: 142-148; Brügger 2002a: 21). The focus of the present article is mainly on the
immediately perceptible layer, but obviously one can advantageously involve the
non-perceptible layers.
5) Even if the archived website is preserved in an Internet archive, it may be subject to
continuous re-writings in relation to long-term preservation, for instance when it has
to be moved to another data format.
6) Due to the digital alphabet, it is possible to some extent to compare archived websites
by automatic means (at least on the non-perceptible level).

165
The Archived WeBsiTe ANd WeBsiTe Philology
7) It is possible that two exactly identical versions of the same website can exist in
two different archives (most likely small, non-complicated websites that are rarely
updated).
Bearing in mind the general considerations mentioned above, we shall now go into
more detail about some of the possible methods and rules that can guide the work of the
website philologist when he/she needs to say something consistent about the differences
and similarities of several versions.23
Source Criticism
First of all, the website philologist must keep in mind the possible existence of different
archived versions of the same web activity. First, one has to realize that the version one
has found is not necessarily the only existing version, second, that one should try to
trace other versions (in the same or in other archives), and third, that the more versions
that are available, the more likely it is that one can determine how close the different
versions are to the website that was once on the live web – and the reverse: the fewer,
the more difcult.
Navigating and Examining
An archived website will very often be faulty and defective. However, the faults and
deciencies are not always immediately visible, but are only gradually revealed as one
uses the archived website. Thus, use and trouble-shooting coincide, and trouble-shooting
is often an ongoing process, which means that very often one has to make use of an
archived website in other ways than a website on the live web.
Some of the frequent faults and deciencies are: rst, that the link structure can
be broken or defective in other ways, second, that textual elements or functions are
missing.24 In the rst case, we have to navigate differently; in the second, we must
examine the archived website using unusual means.
Navigating
In the case of a defective link structure, the following phenomena are often encountered,
each in their own way forcing us to navigate with inventiveness and care.
a) The link does not work, but the link target is in fact in the archive, which is why what
has actually been archived is not always found (this error is often encountered in
menu items, which might cause the user to believe that the link target – e.g. a specic
sub-site/web page – is not part of the archived version). The following alternative
methods of navigating might be used: use a sitemap if possible, make detours (a sub-
site/web page is often linked to from more than one web page), click on anything,
shift between text and graphic version (in older material), cut off the URL-address
from behind (e.g. http://www.dr.dk/Nyheder/Temaer/Oevrige_temaer to http://www.
dr.dk/Nyheder/Temaer to http://www.dr.dk/Nyheder).
b) The link is working, but the link target has not been archived, which can have the fol-
lowing consequences: either we are actually informed that the link target is not in the
archive (this is the case in the Danish national Internet archive netarkivet.dk), or the
link takes us to the correct URL but from another point in time, for instance to another
day, or to the live web (this is the case with the archive archive.org, and the ‘distance’

166
Niels Brügger
between the versions can be up to several years). I use the term ‘inconsistent linking’
for this last phenomenon, and it is not only due to errors in the archived version itself,
but is rather a result of the version being part of an archive with porous boundaries
between the versions or to the live Internet. Thus, inconsistent linking adds a new
dimension to the unreliability of sources.25 In the rst case, nothing is found because
nothing was archived, and there is not much to be done about it; in the second case,
you nd something that is not part of the version from which you departed, and here
we have to navigate with great caution (for instance in the archive archive.org it is
not indicated very clearly that the link is inconsistent, as only the URL-address in the
archive changes, e.g. from http://web.archive.org/web/20000228123340/http://www.
yahoo.com to http://web.archive.org/web/20000229135040/http://www.yahoo.com/
(and inconsistent text is not marked).26
Examining
An archived website without missing elements or functions is very much the exception.
However, valuable information about missing elements is often hidden in the archived
website, thus making it possible to extract relevant information despite the fact that the
information is not immediately manifest. The following three methods can be used to
gain knowledge of missing elements (of the fact that an element is missing as well as
what the element was).
a) There is often a visual mark in the archived website, showing that an element – an
image, a ash – is missing: Either the name of the le or le type appears, or a mouse-
over dialog box communicates the le name or the function that should have been
performed.
b) The source code can reveal a great deal about what should have been displayed on
the screen or why it looks ‘strange’ (for instance because a style sheet is missing).
c) When it comes to missing web pages or sub-sites, the site map or the menu items can
indicate what might be lacking.
Thus, in a number of ways, the archived website itself communicates its own faults and
deciencies in a somewhat systematic and predictable manner (I use the term ‘marked
absence’ for this phenomenon).
However, we cannot rely on these kinds of marked absences. Sometimes an element is
just absent, without any ‘explanation’ and often in an inconsistent manner (for instance
an overall graphical element can be absent on, say, the front page, but present on a web
page in a sub-site). Again, the above-mentioned click-on-anything method might come
in handy.
All in all, with regard to decient link structures and missing elements or functions,
the overall method is: click on anything, use the source code, and examine every corner
of the archived website, even if it appears useless at rst glance.
Evaluating Probabilities
As mentioned above, the website philologist can only determine with various degrees
of probability what a website or a given textual element actually looked like on the live
web – either in terms of the element being present/not present at all, or in terms of the

167
The Archived WeBsiTe ANd WeBsiTe Philology
‘content’ of the element, that is, what is present – and he/she therefore has to make do
with relative comparisons. However, the nature of these relative comparisons is to be
described in more detail. This will occur, rst, by suggesting how relative comparisons
can be performed, and second, with a view to guiding and improving these comparisons,
by discussing the relevance of involving considerations of the proximity in time and
space of the versions, the expected speed of change of the textual elements on the web
page, as well as knowledge of the types of texts on the web page, of genre characteristics
and of typical websites from the period in question.
The Least Decient Version as Original
In light of the absence of an original, one way of getting as close to having an ‘original’
for the comparisons as possible is to use the least decient version as the original (by
‘original’ is not meant ‘as it was on the web’, but just the most complete of the available
versions). Thus, one could proceed in the following manner:
1. Evaluate the different available versions with regard to quality: Are all textual ele-
ments present, do they work properly, are all sub-sites present, etc.?
2. Choose the most complete as the point of departure for the comparisons.
3. Compare this version to the other versions in order to establish differences and si-
milarities; the comparisons can either be in toto (if the similarities dominate) or be
focused on specic points (the textual elements, the number of sub-sites, etc.) if the
differences dominate, in order to avoid too complex a comparison.
Proximity in Time and Space
When we are to compare several web pages to each other, we could use the following
ve-step method, where proximity in time and space of the different archived versions
is prioritized. We should compare the textual elements on the individual web page to
other web pages from:
1. the same version
2. another version from the same day in the same archive
3. another version from the same day in another archive
4. another version from an earlier/later day in the same archive
5. another version from an earlier/later day in another archive
The steps are prioritized from top to bottom, indicating that we lower the probability
the more we are forced to move down the list of steps, thus assuming the following
rule: The closer the compared versions are to each other in terms of time (from same
version to same day to earlier/later day) and space (same version, same archive, another
archive), the greater the possibility of rendering probable what a given textual element
actually looked like on the live web.27 At rst sight, this rule seems to conict with the
requirement mentioned earlier, namely that we should try to trace other versions, and
that the more versions we have available, the more likely it is that we can determine
how close the different versions are to the website on the live web. This is still true, but
the rule of proximity primarily applies when we actually compare the different available
versions, if we have been able to locate more than one.

168
Niels Brügger
Speed of Change
Based on the assumption that the various textual elements on a web page change at dif-
ferent speeds with regard to both position and content, one can place their expected speed
of change within a given period (e.g., a day) on a continuum with two poles: stability and
high-frequency changes. If the same textual element is present at the same position on
more than one of the web pages of the website (or sub-site), it can be considered a stable
element. And if an element is specic to an individual web page, it is more likely to be
changing at a high-frequency. These considerations can be a cause for formulating the
following rule: The more stable an element is, the greater the increase in the possibility
of rendering a probable picture of what it actually looked like on the live web.
Types of Texts
Moreover, a web page can be considered a signifying unit, consisting of a complicated
system of texts and paratexts, paratexts being the small textual elements that serve as a
threshold to the text itself, either on the same page or on others – a menu item, a head-
line, interposed excerpts, a footer, ‘bread crumbs’ that indicate the position of a web page
on a website, etc.28 And these paratexts can form a network on either a local, ‘regional’,
or global scale on a website: For instance, the main menu items that are present on all
(or at least the majority) of web pages are global paratexts; some menu items are only
present on the web pages of a sub-site, which is why they can be considered regional
paratexts, and nally the local paratexts – e.g., specic headlines – are not present on
more than one web page (cf. Fig. 4).
Following these lines of thought, we might assume the following rule: As we move
from the global paratexts of the website via the regional and local paratexts to the text
itself, we also move from stable to presumably high-frequency changing textual ele-
ments, thus decreasing the possibility of rendering probable pictures of what the element
actually looked like on the live web.
However, concerning the texts themselves (and not the paratexts), different expected
speeds of change also seem to be at work, depending on the type of content. A descrip-
tion of a television programme aired a month ago is probably stable; a news item that
is continuously replaced or moved further down the web page can be characterized as
mezzo stable, and text in debates or chat is supposedly changing at a high frequency.
Finally, one can point to other elements of expression that are presumably relatively
stable, such as layout, backgrounds, typography, etc.
Genre Characteristics
The question of expected speed of change also applies to the entire website (or sub-sites)
in terms of genre, because a website can as such be positioned on the continuum between
stable and high-frequency changes. A few examples can illustrate this. A sub-site with
games, etc., for small children is likely to be relatively stable. A sub-site made in relation
to a weekly television serial is probably stable when the serial is not being aired, chan-
ging once a week when it is being aired. A news sub-site can, on the one hand, change
rapidly, but on the other hand it can also have thematic sections (about elections, wars,
etc.) that can vary between frequent changes as the event takes place, relative stability
as it fades out, and the sub-site can end up being totally unchanged when the theme has
ended. And nally, a sub-site with debate is likely to be characterized by high-frequency

169
The Archived WeBsiTe ANd WeBsiTe Philology
changes. We can therefore formulate the following rule: A given textual element is less
likely to have been changed if it is found on a website or sub-site that is supposedly
relatively stable in terms of genre.
Characteristics Typical of the Period
Finally, it can be relevant to involve knowledge of typical websites from the period of
time in question. This could involve, for instance, knowledge about the characteristic
ways of constructing web pages in general, what menu items, navigation features or
other text items may look like and how they may be positioned, use of the paratextual
system, etc.
A Constellation of Indicators
When we are to evaluate the probabilities of what a given textual element actually looked
like on the live web, we can use the above-mentioned phenomena as a set of indicators.
One by one, they can indicate whether the probability is high or low; for instance the
probability is relatively high if we can compare web pages from the same version, and
global paratexts
regional paratexts
local paratexts
text
Figure 4. Texts and Paratexts on the Website

170
Niels Brügger
correspondingly low if we have to compare web pages from versions from a different
day in a different archive, etc. (cf. Fig. 5).
Figure 5. Indicators of High and Low Probabilities
Proximity between versions Distance between versions
Stable textual element High-frequency changes in textual elements
Stable type of text High-frequency changes in type of text
Stable sub-site genre High-frequency changes in sub-site genre
Knowledge of period No knowledge of period
probability
high low
Moreover, these indicators also interact. Thus, the probability is relatively high that a
given element actually appeared on the live web as it is in the archive if, for instance,
the same element is present at the same position in several versions from the same day
in the same archive, and if it is a supposedly stable type of text on a supposedly stable
sub-site genre, and if this is supported by knowledge of typical websites from the period.
And in between the constellations of indicators that are either clearly high or low in
terms of probability, we nd a variety of conicting intermediate forms where the more
the indicators point in different directions, the more blurred the picture becomes.
What has been said thus far about the evaluation of probabilities can lead to for-
mulation of the following general rule: the more indicators that point towards a high
probability, the higher the probability becomes in general, and vice versa. Thus, using a
method that aims at exposing the constellation of the indicators mentioned above enables
us to be more precise when evaluating the probabilities of what a given textual element
actually looked like on the live web.29
Comments and References
In order to facilitate the evaluation of different versions of archived websites, it is recom-
mended that one use the variety of information that in a number of ways ‘comments’ on
the archived website. The following types of information can be of great help.
a) The date and time of the archiving (can be part of the le name or the URL if the
archive is accessible on the Internet (e.g., archive.org)).
b) Supplementary documents, for instance a plan for the archiving process, a log book
or log le (either manually created or automatically generated by the harvesting soft-
ware); these kinds of documents may contain information about the software used,
the starting/ending time and the intervals of the archiving, the starting URLs, the
parameters set, errors encountered, whether the archiving is part of a corpus, etc.
c) Other general information, for instance that a given version is part of a selective
harvesting that has been performed several times a day.
d) Information about how the material has been integrated into the archive, be it by
archiving from the Internet or by delivery from the producer.

171
The Archived WeBsiTe ANd WeBsiTe Philology
The more of these comments that are available, the easier it is for the website scholar to
evaluate a version and to explain faults and deciencies among versions. Besides, this
kind of information can turn out to be very important, at least from the scholar’s point
of view, as existing archives become integrated with one another, for instance when
material from the producers or the scholar who has archived a website as an object of
study is delivered to a national Internet archive (cf. Brügger 2007a: 10).
This information will also make it easier to make unambiguous references to an
archived website. In this regard it is suggested:
• That an archived website be referenced as it has been observed in a given archive, for
instance: "www.dr.dk, 30 January 2005, 02:32, as seen in archive.org, 23 February
2007, http://web.archive.org/web/20050130023236/http://dr.dk/”
• That the reference always be as precise as possible, preferably referring to a specic
web page (in the example above, the web page is the ‘front page’ of www.dr.dk)
Conclusion
As a new eld, website history – and any kind of history writing that involves an ar-
chived website – is still in its infancy, but if it is to become an integrated part of media
history as well as of Internet history in general, we must discuss how the archived
website can be used as a historical document, and what a website archive should look
like if it is to be as useful as possible for media scholars.
To summarize and conclude, it is argued here that any archived website is a re-
construction that does not exist in a stable form before the act of archiving, but is only
created through the archiving process on the basis of web elements – both harvested
and delivered – that are characterized by a variety of dynamics and temporal logics.
Thus, the archived website stands apart from other archived media types, and it chal-
lenges the media historian in new ways, with regard to both the creation and the use of
archived websites.
The media scholar who sets out to archive a website has to take into consideration,
rst, that the processes of research and archiving are closely connected, and second, that
archiving a website has to be accompanied by methodological deliberations. In other
words: What one wishes to examine in a later analysis should to a certain degree already
be anticipated at the time of archiving.
And when the media scholar wishes to use archived websites as sources, he/she has
to bear in mind that if several archived copies of a given website exist from the same
date, they are very likely to be different from one another, and that he/she cannot ex-
pect to nd an original in the form of the website as it actually looked on the Internet
at a given time. Therefore, the archived website has to be treated differently than other
well-known types of documents, even – or maybe especially – the website on the live
web. The present article has attempted to outline some of the methodological principles,
rules and recommendations that can help the media scholar in this task, based on the as-
sumption that a number of the fundamental concepts and questions in ‘classical’ textual
philology are relevant to a critical textual philology of the website, and that they have
to be brought into line with the specicity of the archived website.
Finally, it should be stressed that the problems of the archived website are not only
relevant to website history, as in practice most website studies are based on some kind

172
Niels Brügger
of archiving of the website with a view to preserving a stable analytical object, except
for studies of the live web. Therefore, what has been argued about the special charac-
teristics of the archived website is not limited to website history, but is in fact broader
in scope, aimed at website studies in general.
The present work has made only a small contribution to the foundation of web-
site history by putting some of the fundamental methodological issues on the agenda.
Obviously, more work is needed in the future to elaborate and rene the theoretical
considerations as well as the methodological principles, rules, and recommendations
outlined above.
Notes
1. For an overview of some of the theoretical and methodological challenges and discussions within
media history, see Bondebjerg 2002; Brügger 2002b; Dahl 2002; Dahl 2004; Djerf-Pierre 2002;
Godfrey 2006; Jensen 2002; Salokanges 2002; Schanze 2001; Snickars 2006; Startt & Sloan
1989.
2. For instance, an adequate description of the website cannot be based on well-known text-related
analytical concepts such as newspaper article, work, programme etc., just as reception studies
must be based on new ideas of interaction. However, these re-evaluations are not only of rel-
evance to website history, but also to website analysis in general.
3. Some of the rst major contributors to Internet and web history are Poole 2005; Abbate 2000;
Naughton 2002; Gillies & Cailliau 2000; Henderson 2002; Hauben & Hauben 1997.
4. An intermediate analytical level between the Web and the website could be what Schneider &
Foot describe as ’the web sphere’: ”a set of dynamically dened digital resources spanning mul-
tiple Web sites deemed relevant or related to a central event, concept, or theme” (Schneider &
Foot 2006: 20, cf. also 20-21, 27-35). In web sphere analysis, the focus is less on the individual
web site than on the theme that unites various websites (one could maintain that it is a way of
conceptualizing web activities that is similar to the archiving strategy of ’event harvesting’; cf.
Brügger et al. 2003: 6-7).
5. The underlying and fundamental question of what is understood by ’website’ will not be dis-
cussed in this article. It should just be mentioned that no matter which criteria one uses to
delimit a specic website, historical studies of websites have added a new dimension to ‘website
studies’, because it can be difcult to decide how the website should be delimited over time: Did
it have the same boundaries two, ve and ten years ago? For a denition of what can be under-
stood by ’website’ see Brügger 2007a: 84-87, Brügger 2009 (forthcoming), and Brügger 2008b
(forthcoming).
6. For a general introduction to and discussion of web archiving, see Brügger 2008b (forthcom-
ing).
7. These six ways of integrating web material into a web archive are claried and discussed in more
detail in Brügger 2007b: 3-6.
8. Other subjective elements are related to long-term preservation – for instance, when the mate-
rial has to be moved to another data format or to another media type (e.g., analogue to digital).
For a discussion of the problems of long-term preservation of web content, see Masanès 2006:
177-199.
9. Cf. Brown 2006, Brügger et al. 2003, Brügger 2005, Masanès 2006, as well as papers and pro-
ceedings from the International Web Archiving Workshop (IWAW, organized since 2001, see
www.iwaw.net). Cf. also reports and other activities at the International Internet Preservation
Consortium (http://netpreserve.org).
10. The problem of updating is discussed in more detail in relation to the front pages of newspaper
websites in Falkenberg 2006: 8-9. What changes is the news item in toto, but on both a macro
level (logo, overall structure and layout etc.) and a micro level (the actual text of the news item),
newspaper websites tend to be rather static, as is evident, for instance, in the fact that the text of
the news item is rarely changed once it has been published.
11. This also applies to my book Archiving Websites. General Considerations and Strategies (Brüg-
ger 2005), which only very briey refers to delivery-related problems in the rst chapter (p. 12).
To my knowledge, one of the rst experiments with delivery from the producer to an Internet
archive was carried out in 2001 in the ‘netarkivet.dk’ research project, the precursor to the na-

173
The Archived WeBsiTe ANd WeBsiTe Philology
tional Danish Internet archive “netarkivet.dk”, established in 2005 (cf. Brügger et al. 2003: 19,
36). This project has been followed (spring 2007) by another research project: “Method study
of the integration of internet material directly received from producers” (cf. the section ‘Test of
versions in existing web archives’ in the present article).
12. One of the lessons drawn from the ’netarkivet.dk’ research project (2001) is that it is often nec-
essary to establish an identical copy of the running environment (cf. Brügger et al. 2003: 19).
13. Copyright and economic problems are other problems that could be pointed out (cf. Brügger
2005: 12).
14. Evidence of the former is available on the Air-l-list from May 2007: ”I just started to use Zotero
[...]. Now I want to share my grabs with my colleagues. I think they have not yet attacked the
sharing of saved pages and annotations. Or do I miss something?” (Posted by Frank Thomas,
10 May). An example of the latter is the retrospective collection of Danish websites that the Na-
tional Danish Internet Archive ”netarkivet.dk” initiated in 2005, aimed at media scholars.
15. The problems related to the montage of harvested web elements are discussed in relation to mi-
cro archiving in Brügger 2005: 39-62. Several of these considerations also apply to the delivery
of web material.
16. For instance, a substantial part of my own web archive, begun in 2000, is made with Internet
Explorer for Mac (creating WAFF les), which is why this material can only be displayed with
this conguration. Besides, the archive also contains a number of screen shots and screen mov-
ies (i.e. les that are neither HTML nor ARC).
17. The overall aim of the project is to write the history of the rst ten years of the Danish Broadcast-
ing Corporation’s (DR) website, i.e. the period from 1996 to 2006 (dr.dk has been the biggest
Danish website for some years). The project is supported by the Danish Research Council for
the Humanities in 2007-8. The development of the project website www.drdk.dk is supported
by the ”Knowledge Society” research priority area of the Faculty of Humanities, University of
Aarhus.
18. The aim of the pilot project is to analyze the archiving and research problems and possibili-
ties connected with the collection of Internet material from the producer and the integration
of these in the Danish National Internet Archive ”netarkivet.dk” (”netarkivet.dk” is run by the
Royal Library and the State and University library). The pilot project is being carried out in col-
laboration with the Danish State and University Library/”netarkivet.dk” and is supported by the
research foundation of the Ministry of Culture. The technical results of the test with regard to
the transformation of non-harvested web data into ARC les are discussed in Andersen 2007.
The general results of the test are discussed in Brügger 2008a (forthcoming).
19. As regards the website, the word ’textual’ is meant to refer to coherent units or forms of expres-
sion such as written words, still images, moving images, and sound. Therefore, in the following,
‘text’ is understood in a broad sense and is not merely limited to written language. (cf. also Brüg-
ger 2007a: 75).
20. Cf. the section ”Finding, collecting and preserving websites” above. For an elaboration of the
term ’media materiality’, see Brügger 2002a; Brügger 2002b: 44-52.
21. For a short introduction to and discussion of the different traditions within textual philology,
see Cerquiglini 1999: 46-71; for a discussion of some recent tendencies, see the contributions in
Tervooren & Wenzel 1997. In addition, it should be noted that when the theoretical and meth-
odological inuence between Internet theories and ’classical’ textual philology is on the agenda,
focus is mostly on how Internet theories in the broadest sense, for instance hypertext theories,
could be used to enhance digital editions of medieval and other manuscripts (e.g., Stolz 2003;
Carlquist 2004). The present paper sets out to discuss possible inuences in the opposite direc-
tion, from ’classical’ textual philology to Internet studies.
22. That the printed copy is called ’authoritative’ is understood in a merely technical way, as all later
printed copies of a given edition are alike (cf. also Müller 2006: 187-188). This is a major dif-
ference to what characterized the scribe culture of manuscripts, where errors, corruptions and
variations were the order of the day. However, errors and corruptions are also a part of print cul-
ture, but in contrast to scribe cultures, the number of errors decreases over the years and, most
importantly, the printers, editors, and reading communities are conscious of the existence of
different copies, a fact of which both the continuous publication of errata and the ’standardiza-
tion of errors’ (cf. Eisenstein 1996: 51, 59) as well as the actual establishment and development
of a critical textual philology are evidence. But although erroneous or variant printed editions
can be issued – for instance editions with misspellings, or editions based upon different sources
and methodological approaches – once a new printed edition is published, all later copies of it
are identical, thus making it ’authoritative’.

174
Niels Brügger
23. It should be mentioned that the methods and rules outlined in the following are not all relevant
in relation to all types of archived websites.
24. A textual element is understood as a dened coherent textual unit, composed of one of the fol-
lowing four formats of expression: written words (or the like), still images, moving images, or
sounds. Textual elements could, for instance, be a headline and the body text (writing), a photo-
graph (image), a banner ad or a news story from television (moving images), or a piece of music
(sound) (cf. Brügger 2007b: 84-85).
25. A special type of inconsistent linking is what could be termed ’inconsistent text’, namely the fact
that a textual element on an archived web page contains commands that are set to get the cur-
rent news headlines, weather forecast or the like from a web server (cf. the section ”Archiving
from the net” above). In the archived version, this information will be retrieved automatically
from the live web (if the URL still exists), but without making note of the fact. Basically, we are
dealing with an inconsistent link, but the content of the link target is seamlessly merged into
the text of the archived webpage, which makes the text as such inconsistent, a fact that is not
brought to the user’s attention.
26. This method of inconsistent linking in archive.org is part of a deliberate policy, cf. “Not every
date for every site archived is 100% complete. When you are surng an incomplete archived site
the Wayback Machine will grab the closest available date to the one you are in for the links that
are missing. In the event that we do not have the link archived at all, the Wayback Machine will
look for the link on the live web and grab it if available” (http://www.archive.org/about/faqs.
php#202, accessed 12 July 2007).
27. This method presupposes that one has more than one web page from the website in question,
which is normally the case if one has access to an archived website. If one has only one web page
from a given website (e.g., a pdf le or the like) and neither (parts of) the whole website nor ver-
sions from the immediately surrounding days, one is in an unfortunate position.
28. The history of paratexts is discussed by Gérard Genette in Genette 1997, mostly with regard
to the printed novel. Danish scholar Finn Frandsen has outlined a development of Genette’s
insights in order to adapt them to the newspaper, see Frandsen 1991 and 1992. Cf. also Brügger
2007a: 86-87 for a brief discussion of paratexts in relation to the website.
29. This method focuses on the textual elements of the individual web page, but it will probably also
be useful in determining the structure of a website, i.e. the presence of sub-sites.
References
Abbate, J. (2000) Inventing the Internet. Cambridge, Mass.: The MIT Press.
Andersen, B. (2006) ‘Integration of Non-harvested Web Data into an Existing Web Archive’, http://netarkivet.
dk/publikationer/IntegrationOfDeliveredData.pdf, November 2007.
Bondebjerg, I. (2002) ’Scandinavian Media Histories’, Nordicom Information, 24, 2-3.
Brown, A. (2006) Archiving Websites. A Practical Guide for Information Management Professionals. London:
Facet Publishing.
Brügger, N. (2002a) ‘Does the Materiality of the Internet Matter?’, in N. Brügger, H. Bødker (eds.) The Internet
and Society? Questioning Answers and Answering Questions, papers from The Centre for Internet Re-
search no. 5. Aarhus: The Centre for Internet Research, pp. 13-22.
Brügger, N. (2002b) ‘Theoretical Reections on Media and Media History’, in N. Brügger, S. Kolstrup (eds.)
Media History: Theories, Methods, Analysis. Aarhus: Aarhus University Press.
Brügger, N. et al. (2003) Experiences and Conclusions from a Pilot Study: Web Archiving of the District and
County Elections 2001. Final Report for The Pilot Project ‘netarkivet.dk’, netarkivet.dk, Copenhagen,
http://netarkivet.dk/publikationer/webark-nal-rapport-2003.pdf (retrieved May 2007)
Brügger, N. (2005) Archiving Websites. General Considerations and Strategies. Aarhus: The Centre for Internet
Research.
Brügger, N. (2007a) ‘The Website as Unit of Analysis? Bolter and Manovich Revisited’, in A. Fetveit, G.B.
Stald (eds.) ‘Digital Aesthetics and Communication’, Northern Lights: Film and Media Studies Yearbook,
vol. 5, Bristol: Intellect, pp. 75-88.
Brügger, N. (2007b) ‘Website History: Theoretical and Methodological Problems in an Emerging Field’, paper
presented at The Association of Internet Researchers Conference Internet Research 8.0: Let’s Play, Van-
couver, 18-20 October 2007.
Brügger, N. (2008a) (forthcoming) Archived Websites between Copies and Versions: Test of Versions in Existing
Web Archives. Papers from The Centre for Internet Research: Aarhus: The Centre for Internet Research.

175
The Archived WeBsiTe ANd WeBsiTe Philology
Brügger, N. 2008b (forthcoming) ‘Web Archiving – between Past, Present and Future’, in R. Burnett, M. Con-
salvo and C. Ess (eds.) Blackwell Companion to Internet Studies. Oxford: Blackwell.
Brügger, N. (2009) (in press) ’Website History and the Website as an Object of Study, New Media & Society,
11: 1-2.
Carlquist, J. (2004) ‘Medieval Manuscripts, Hypertext and Reading. Visions of Digital Editions’, Literary and
Linguistic Computing, 19, 1, pp. 105-118.
Cerquiglini, B. (1999) In Praise of the Variant. A Critical History of Philology. Baltimore/London: The John
Hopkins University Press.
Dahl, H.F. (2002) ‘The Challenges of Media History’, Nordicom Information, 24, 2-3.
Dahl, H.F. (2004) Mediehistorie. Historisk metode i mediefaget. Oslo: Damm.
Djerf-Pierre, M. (2002) ‘The Logic and Practice of Writing Journalism History’, Nordicom Information, 24,
2-3.
Eisenstein, E.L. (1996) [1993/1983] The Printing Revolution in Early Modern Europe. Cambridge: Cambridge
University Press.
Falkenberg, V. (2006) ’Metodologiske utfordringer ved arkivering av nettaviser’, paper presented at Mediefors-
kerlagets Årskonferanse, Bergen 19-20 October.
Finnemann, N.O. (1999) ‘Modernity Modernised – the Cultural Impact of Computerisation’, in P.A. Mayer(ed.)
Computer, Media and Communication. Oxford: Oxford University Press, pp. 141-159
Frandsen, F. (1991) ’Avisens paratekst – et nyt område for medieforskningen’, Mediekultur, 16, pp. 79-97.
Frandsen, F. (1992) ‘News Discourse: The Paratextual Structure of News Texts’, in A.-C. Lindeberg, N.E.
Enkvist, K. Wikberg (eds.) Nordic Research on Text and Discourse. Åbo: Åbo Academy Press, pp. 147-
59.
Genette, G. 1997. Paratexts: Thresholds of Interpretation, Cambridge: Cambridge University Press.
Gillies, J. & R. Cailliau. (2000) How the Web was Born. The Story of the World Wide Web. Oxford: Oxford
University Press.
Godfrey, D.G. (2006) Methods of Historical Analysis in Electronic Media. Mahwah, NJ: Erlbaum.
Hauben, M. & R. Hauben (1997). Netizens. On the History and Impact of Usenet and the Internet. Washington:
IEEE Computer Society.
Henderson, H. (2002) Pioneers of the Internet. San Diego: Lucent Books.
Jensen, K.B. (2002) ‘From Media History to Communication History’, Nordicom Information, 24, 2-3.
Masanès, J. (ed.) (2006) Web Archiving. Berlin: Springer.
Müller, J.-D. (2006) [2002] ‘The Body of the Book. The Media Transition from Manuscript to Print’, in D.
Finkelstein, A. McCleery (eds.) The Book History Reader. London/New York: Routledge, pp. 182-189.
Naughton, J. (2002) [1999] A Brief History of the Future. The Origins of the Internet. London: Phoenix.
Poole, H.W. (red.). (2005) The Internet. A Historical Encyclopedia. Santa Barbara: ABC/Clio.
Salokangas, R. (2002) ‘Media History becomes Communication History’, Nordicom Information, 24, 2-3.
Schanze, H. (2001) Handbuch der Mediengeschichte. Stuttgart: Kröner.
Schneider, S.M. & K.A. Foot. (2004) ‘The Web as an Object of Study’, New Media & Society, 6, 1.
Schneider, S.M. & K.A. Foot. (2006) Web Campaigning. Cambridge, Mass.: The MIT Press.
Snickars, P. (2006) ’Om ny och gammal mediehistoria’, Nordicom Information, 28, 1.
Startt, J.D & D. Sloan (1989) Historical Methods in Mass Communication. Hillsdale: Erlbaum,
Stolz, M. (2003) ‘New Philology and New Phylogeny: Aspects of a Critical Electronic Edition of Wolfram’s
Parzival’, Literary and Linguistic Computing, 18, 2, pp. 105-118.
Tervooren, H. & H. Wenzel (eds.) (1997) Philologie als Textwissenschaft. Alte und neue Horizonte (Zeitschrift
für deutsche Philologie, 116. band, Sonderheft) Berlin/Bielefeld/München: Erich Schmidt Verlag.