[tei-council] my letter in email form as requested by Martin Holmes

[Martin Mueller]

Here is the text of my letter in email form, as requested by Martin Holmes.
Broadly speaking this adds up to an anecdote in interoperability.

To members of the TEI-C Board and Council
>From Martin Mueller, chair, TEI-C Board
 
August 4, 2011
 
Dear colleague,
 
What follows is a long letter to you containing reflections and suggestions
about goals and directions for the TEI-C in the years to come. This is not a
set of formal proposals but a starting point for discussion. If the letter
helps to frame that discussion it will have served its purpose.
 
Since I have never held office in the TEI-C and am in many ways a newcomer
and outsider, I would like to turn that handicap into an opportunity and
speak my mind. If some of what I say turns out to be just silly, it will be
easy to ignore. If some of it is helpful, there may be a use for it for a
conversation with our members --and more particularly the research libraries
that are our main source of support--about what we can or should do for and
together with them.
 
This is a rambling piece in which I think through various issues relating to
the TEI that strike me as important and that at some point before the
Würzburg meeting I want to raise with the membership in a more formal
document. I hope that your responses will help separate the wheat from the
chaff, not to speak of matters that I have overlooked.  Questions that I
return to, perhaps not in the best order concern
 
1.   the ways in which the TEI has succeeded and failed,
2.   the lack of tools that help scholarly end users make use of the added
value that is in principle created by structural encoding,
3.   whether the TEI should have a seat at the table of ³Big Data,²
4.   what role it can plays in ³research data lifecycle management,²
5.   what role it might play in collaborative transcription of manuscripts
by amateur scholars
6.   how to address its hybrid role as a standards body,  a scholarly
society, and a membership association
7.   how to depend less on the support of a few institutions and move
towards more broadly based support of more members with a lower fee
structure
8.   how to establish a proper balance between financial contributions from
North America and Europe
9.   whether the TEI-C is a top-heavy organization relative to the size of
its membership and the scope of its operations
 
I do not think that this letter contains any confidential data, and you may
share it as you see fit.  I want to stress, however, that in this letter I
speak entirely for myself, talking about things as I see them after
listening  to you and talking with you, as well as with various outsiders,
about TEI matters.
TEI as an enabling technology for the world of letters
At the heart of the TEI is a markup language designed to model in a digital
environment texts that originated in a world of print and manuscripts. The
TEI operates in a world of "letters," an old-fashioned but useful word, and
it is an enabling technology for text-centric research in history,
linguistics, literature, philosophy, religion, and cognate disciplines. TEI
encoding can certainly be used for other purposes, but unless it is accepted
as a critical or important enabling technology by historians, linguists,
literary scholars, philosophers, theologians, etc., there is not much point
in having it in the first place.
 
TEI markup uses XML, which is a business technology. There is nothing wrong
with that. The Greek alphabet was adapted around 800 BCE from a Semitic
alphabet, almost certainly because Phoenician and Greek traders wanted to
keep track of what they were doing. But the earliest surviving piece of
Greek writing is a hexametric line on a vase that proclaims to be Nestor's
drinking cup and speaks of the aphrodisiac powers it bestows on its owner. I
believe that T. S. Eliot typed the Wasteland  on a machine that at the time
was still a relatively new piece of business technology. Commerce and
culture or "Besitz und Bildung" are very old and close cousins.
 
TEI is a scholarly technology, and its "value proposition" --an ugly but
useful term from the world of business-- consists in the promise for deeper
or more extensive inquiries that properly encoded TEI texts offer to
text-centric scholars. By comparison HTML is not a scholarly technology,
although it certainly can be used for scholarly purposes and has been used
so with varying success.
 
Because the TEI is a dialect of XML it is bound by two fundamental
assumptions of that language. The first of these is the assumption that you
can neatly separate the structure of a document from the manner of its
display. The second is that every text is an "ordered hierarchy of content
objects" in the sense that a single hierarchical model accounts for all
structural aspects of a text. Most text-centric scholars will have problems
with these guiding assumptions. On the other hand, both of them work often
enough to be heuristically useful. In Wallace Stevens' poem "Connoisseurs of
Chaos" there is the wonderful line "The squirming facts exceed the squamous
mind." It is a good line to remember when dealing with just about any
attempt to contain human language, whether written or spoken, within strict
rules. 
Success and failure
>From one perspective, the TEI has exceeded expectations. Virtually all
digital editions of primary texts with any claim to scholarly standards use
it. TEI is the lingua franca of digital scholarly editing on a global basis.
You find it in editions of Buddhist sutras, New Zealand and Pacific island
texts, Greek inscriptions, French manuscripts of the Roman de la rose, the
Hengwrt manuscript of the Canterbury Tales, slave narratives of the American
South, or the historical records of the State Department. TEI has been used
in all the large-scale university- library-based digitization projects of
primary texts at Indiana, Michigan, North Carolina, Virginia, and the
Library of Congress. The same is true of European encoding projects.
 
That is the success story. But now consider a thought experiment where you
ask the chairs of history, literature, linguistics, philosophy, and religion
departments of the world's 100 top universities to write a sentence or short
paragraph about the TEI. These would be very short sentences or paragraphs.
The one message you would not get from them is the recognition that the TEI
offers an important enabling technology for work in their disciplines.
Results would be a little better (but not much) if you asked the chief
librarians of those universities or their technology officers. You would get
a little more knowledge but no ringing endorsement of the TEI as a key
technology for some disciplines. You'd get the eye-rolling and
shoulder-shrugging that comes with humoring a bright but wayward child. At
least that has been my experience. In a forthcoming piece for the Profession
issue of the MLA Jerry McGann observes that TEI stands for "terra
incognita." The pun may be cruel, but I fear there is something to it.
Decoding the encoded
What accounts for this disconnect and what could or should we do about it?
The TEI is about "encoding." The point of any encoding is decoding at the
other end. A book is an encoding for human readers who decode it. What about
the decoding of TEI encoded documents? There is a value and pleasure
(masochistic at times) for TEI mavens who wrangle recalcitrant textual
phenomena into the corset of a TEI schema. There may be an even deeper
pleasure in cursing the inadequacies of the out-of-the-box schema and
customizing it to a point at which it expresses all or most of the
"squirming things" that do not fit into the "squamous mind." The pleasure of
"difficulté vaincue." I googled the term because of my shaky grasp of French
orthography. At the top of the list comes "Satisfaction, Difficulté Vaincue"
showing a rock climber looking down rather smugly. What would we do without
Flickr and Google? (http://www.flickr.com/photos/45169745@N00/46485696/)
 
But what about the added value of TEI specific encoding for the historian,
linguist, philosopher, literary critic etc.? How can they decode or get at
it, and what does it do for them? The answer is that for the most part they
cannot get at it at all. I remember a conversation with a librarian who said
something like "Oh yes, those TEI texts. We put them through the Lucene
indexer and that's pretty much it." In principle, TEI encoding increases the
query potential of the digital surrogate that is created by it. In practice,
most of that query potential is ignored by the indexing and search software
through which the encoded texts are mediated. Or if it is not ignored, it is
used as instructions for XSLT style sheets to render the XML in HTML. As a
result, the scholarly end users who encounter TEI-encoded texts almost never
encounter them in an environment where they can take advantage of the
distinct affordances of that encoding.
 
You can spend a lot of time explaining to your colleagues in an English
department that it is a wonderful thing for texts to be encoded in TEI
because it offers a much more robust, granular, and flexible way of storing
textual data in digital form. But if what they see is browser-rendered HTML
and if what they search might as well be a plain text file, it is not easy
to persuade them of the value of this robust, granular, and flexible
encoding. It does nothing to help them with their current project. Thus it
is not much of an exaggeration to say that for ordinary scholarly users, TEI
encoded texts right now offer no advantage over plain-text, html, or epub
texts. Nietzsche once exclaimed in exasperation: "Was hilft mir der echte
Text wenn ich ihn nicht verstehe?" or "What use is the true text if I don't
understand it?" One might vary this into "What use is the encoded text if I
cannot decode it?"
 
This is of course not true of the many and mostly small-scale projects that
offer carefully and sometimes exquisitely curated editions of this or that
set of documents. I yield to nobody in my admiration for the philological
and technical ingenuity or beauty of design that has gone into these
projects. But allow me to be quite blunt for a moment and say something that
I have believed strongly ever since I learned about the TEI in the
mid-nineties. While small-scale and lovingly curated projects will have an
important place in the ecosystem of digital textuality, the TEI will become
an increasingly marginal niche player unless it stakes out what scholars and
librarians recognize as a compelling role in the curation and exploration of
the large-scale corpora that are changing the ways in which research is done
in many disciplines.
Big Data and Research Data Lifecycle Management
Oren Sreebny at the University of Chicago wrote a running blog about a
recent Princeton workshop about "research data lifecycle management"
(http://blog.orenblog.org/ <http://blog.orenblog.org/> ). In reading through
it, I was struck by two quotations attributed to Brian Athey, a
bioinformatics researcher from Michigan, in a talk about Big Data 2011:
 
It¹s difficult to incentivize researchers to share data.

 

Agile data integration is an engine that drives discovery.

 
"Interoperability" is the word that negotiates the space between these two
statements. This is an area where TEI practitioners have done less in the
past than they might have and need to do a lot more in the future. The
ability to scan very rapidly across large masses of textual data is a quite
recent phenomenon. It has been dominated in the popular and the scholarly
mind by "Googling" or shallow forms of text retrieval that work across
billions or trillions of words in uncurated and often messy data. In the
lifetime of the TEI, Natural Language Processing tools and methods have made
enormous strides. Thus 1991 the British linguist John Sinclair wrote in the
preface to his Corpus Concordance Collocation:
 
Thirty years ago when this research started it was considered impossible to
process texts of several million words in length. Twenty years ago it was
considered marginally possible but lunatic. Ten years ago it was considered
quite possible but still lunatic. Today it is very popular.

 
That was then. The British National Corpus, a stratified sample of 100
million words of contemporary English, was released in 1994. Over the past
year, Mark Davies at Brigham Young University has released two 400 million
word corpora of Historical American English (1800-) and Contemporary
American English (1990-2010). The Text Creation Partnership, probably the
most ambitious academic transcription enterprise, will by 2015 produce
TEI-encoded digital versions of some 70,000 public domain texts of English
texts before 1800, with a word count of somewhere between 5 and 10 billion
words. Google's n-gram corpus provides limited access to 500 billion words
in some five million books.
 
These changes of scale have attracted the attention of scholars and
librarians in ways that the TEI has not. Ask a modal chief librarian about
the TEI, and s/he hardly knows what to say. Ask her or him about Google
Books or Hathi Trust, and s/he hardly knows how to stop talking. Then there
are the NLP and information retrieval folks who love Big Data and argue that
if you have very large data sets you will get enough 'signal' even if the
noise level is high. That is partly true, but it is also a convenient
fiction because whatever you may want to say about reducing noise level in
large data sets, it is very boring work.
 
Data curation has emerged as an important cross-disciplinary topic in recent
years. In a network-based environment, there has been a lot interest in
user-generated curation. "Crowdsourcing" and "dispersed annotation" are
terms that float in various contexts. The interest is due to a growing
recognition that more data may not be enough. There are lots of things you
may have to do "to" the data before you can do useful things "with" them. If
you follow the rhetoric of the Hathi Trust it is still preoccupied with
"more" and takes a Herodotean delight in pure enumeration. On its home page
you learn that its 9,410, 319 total volumes stretch for 111 miles and weigh
7,653 tons. But behind the scenes there is more interest in data quality and
steps that would increase the interoperability of large textual data sets.
  (http://www.hathitrust.org/home)
 
Is there a role for the TEI in the "research data lifecycle management" of
large primary source materials in the humanities? Not of everything, but of
very large data sets that repay higher levels of curation because such
curation enables inquiries that could not be done otherwise. Corpus
linguists tend to be uninterested in or are openly skeptical about the
structural annotation of text corpora and like to think that "smart"
processing of "dumb" data can give them all they need. They may be right for
their domains, but in the humanities it is unlikely that researchers will
have the programming skills that are common in the NLP community. There is
likely to be a considerable research potential in coarse but consistent
structural annotation that is applied across very large data sets and made
the basis for query tools that can support quite granular queries through
the intersection of coarse criteria.
 
You may not want to do this for millions of books, but within a given
language it may be worth doing for tens of thousands of texts in ways that
start from some "seed corpus" and rely on user-driven and collaborative
curation and augmentation. In recent correspondence, Neil Fraistat and Doug
Reside came up with the acronym CRIPT for "curated repository of important
primary texts" -- an entity that could live as a special collection or inner
ring inside much larger aggregates like Hathi Trust or Google Books. But
while such entities would be smaller by orders of magnitude than the
aggregate of all digitized texts, each of them would be considerably larger
than could be managed by individuals or departmental projects. We are in a
world of changed scale with requirements for robust infrastructure and
highly skilled technical staff that are most easily envisaged as part of
large academic libraries or GenBank like institutes.
 
In such a vision of things the TEI can easily claim a role as a critical
enabling technology for any structural modeling of document structure, and
it is really the only game in town. It can also claim that pieces of such a
vision exist here and there, whether in the French MEET Project, the
large-scale German corpora under development by the Berlin-Brandenburg
Academy and the TextGrid project, or the Text Creation Partnership.
 
What is needed for such a vision to succeed is a collective interest in
query tools, and a much stronger commitment to interoperability across
diverse data sets. By and large, the TEI has followed a "perl" ethos of
reveling in the fact that the same thing can be done in many ways. I
remember a conversation over a decade ago with a young programmer, who told
me that he liked Python better than perl. I had never heard of Python and
asked him why. He said that in Python there was typically only one way of
doing a particular task and that made it easier to write consistent and
reusable code. Much later I came across the Zen of Python and especially
liked #13
 
            There should be one -- and preferably only one -- obvious way to
do it.

 
There are also quite a few occasions of looking at TEI encoding when I have
wistfully remembered "Flat is better than nested" (#5). And I have spent
many hours listening to the tirades of very gifted programmers about the
inconsistency and unnecessary complexity that make TEI texts difficult or
impossible to process. Within the confines of an individual project you can
typically work around such problems. But if you want to work across corpora
the problems very quickly turn into roadblocks. It would help a lot if the
TEI adopted a "Pythonic" ethos and encouraged its practice with various
combinations of preaching and nagging.
The rise of xquery, a turn towards decoding, and the need for
interoperability
Unnecessary divergence and inconsistency --eloquently and often lamented by
Mark Olsen -- are only one reason for the fact that much of TEI encoding is
"lost in translation." Much more important is the fact that the development
of XML aware search engines has lagged behind encoding. Charles Rosen, not a
friend of original instrument performances, once wrote that sonatas like the
Appassionata were written for an instrument that did not exist until thirty
years after Beethoven's death, when Theodore Steinway, now in New York,
added cross-stringing to the cast iron frame that Chickering had pioneered
in Boston in the 1820's. In the XML world we do not yet have query tools
that match the scale and complexity achieved by 30 years of tweaking SQL and
15 years of making SQL work with the Web. But over the past decade xquery
and the emergence of XML databases like eXist or BaseX give considerable
hope for the future. There is also the potential for mimicking some XML
awareness by loading information about some elements as "positional
attributes" into the indexes of Lucene or the CQP query language.
 
These are very promising developments, but if you look at encoding from the
perspective of subsequent decoding by an end user who wants to constrain a
result set by combining various elements you may discover that less is more.
The crossing of even a handful of criteria creates a great variety of
possible searches. 
 
I conclude from these reflections that it might be a good idea for the TEI
to concentrate for a while on the problems and opportunities of encoding and
decoding "at scale." This would shift the focus of attention to a quite
basic set of tags and to the ways in which they could be made to work across
many data and create, so to speak, a TEI-API for query tools. The TextGrid
concept of "Kernkodierung" or "base line encoding" points in that direction.
The point is not to prevent elaboration or complexity where it is necessary,
but build levels of complexity on a pyramidal base to which texts from
different collections can be reduced and which sits considerably above the
word token. Think of a "highest common factor" rather than the lowest common
denominator. 
 
By way of example, a TEI-P5 set of about 60 elements, not counting the
header, is all that is needed to encode the large variety of Early Modern
English texts in the TCP. Making these 60 tags work better, promoting their
consistent use, and explaining to a lay audience how they work and what you
can do with them may for quite a while be a better use of TEI resource than
refining or adding to the current element set. The three percent of
"squirming facts" that wriggle out of any corset will always be of greater
interest to humanists with their innate fondness for the haecceitas of
things. But there is a greater pay-off in the boring work of designing and
maintaining robust and consistent practices for the limited set of elements
that can deal quite adequately with the 97% of cases that do not resist the
squamous mind. 
 
The issues I'm reflecting about here are hardly specific to the TEI. In an
earlier posting to the TEI list I quoted from the Lexus project at the Max
Planck Institute for Psycholinguistics:
 
Lexicography in general is a domain where uniformity and interoperability
have never been the operative words: depending on the purpose and tools used
different formats, structures and terminologies are being adopted, which
makes cross lexica search, merging, linking and comparison an extremely
difficult task. LEXUS is also an attempt at putting an end to this problem.
Being based on the Lexical Markup Framework (LMF), an abstract model for the
creation of customized lexicons defined following the recommendations of the
ISO/TC 37/SC 4 group on the standardization of linguistic terminology, LEXUS
allows on the one hand to create purpose-specific and tailor-made lexica,
and on at same time assures their comparability and interoperability with
other resources. 

 
Whether they will succeed is far from clear. But if the TEI wants to engage
the growing interest in corpus-based inquiry "at scale" it needs to put much
more emphasis on achieving high levels of interoperability for a limited
element set. In terms of a product cycle, we may need to focus more on
implementation and execution of the "Top Sixty" tags than on invention or
design of new things. And marketing the "Top Sixty" with flair and simpler
documentation may pay off as well. In this context it has been instructive
to me to follow the correspondence between Google's Ranjith Unnikrishnan and
several council members about transforming Google books into curatable P5
rough cuts. I draw from it the tentative conclusion that we lack a good
short overview of the TEI. I have a somewhat similar response to the
excellent TEI by Example site, where the documentation is written a little
too much from within. Too much of TEI stuff is written from within.
 
To sum up this part of my letter, it may be a good idea for the TEI in the
coming years to focus on scholarly end users not as encoders of data but as
decoders of already existing TEI texts. This involves finding answers to two
questions:
 
1.   How can we make sure that the benefits of encoded texts are in fact
delivered to the scholarly users for whom these encodings were intended in
the first place?
2.   Should there be a place for the TEI at the Big Data table, and if so,
what should we do to get a place at that table?
 
Agribusiness and organic farming
While I believe that the TEI's engagement with Big Data is important I don't
want to be heard as saying it should be the only game in town. From the
scholarly end user's perspective the critical problem is how to decode the
encoded or unlock the query potential with which encoders have enriched the
source data. This critical problem applies equally at the micro- and
macro-levels, and it may be that good solutions will come from scaling up
small or mid-scale projects. A few months ago, Katherine Rowe, the chair of
English at Bryn Mawr, who has a deep interest in digital tools as a way of
promoting research opportunities for undergraduates, drew my attention to
the work of Robert Binkley, an American historian of the 1930's and leader
in the WPA local history project. In 1935 Binkley wrote a remarkable essay
in the Yale Review called ³New Tools for Men of Letters.²
(http://www.wallandbinkley.com/rcb/articles/newtools-output.html)
A lead sentence in that essay reads: " The new graphic arts devices are, I
believe, capable of working the other way ‹ as implements for a more
decentralized and less professionalized culture, a culture of local
literature and amateur scholarship."
 
I wrote to her:
 
 I looked at Robert Binkley's essay and was as taken with it as you have
been. It's a virtual contemporary of Walter Benjamin's essay on the work of
art in a mechanical age, and in its way it worries about not unrelated
things. The technologies have changed, but the ideological framing has
stayed very much the same. A lot of humanists (and not only humanists) like
to shop at farmers' markets. I think of Google and Hathi Trust as textual
agribusiness. Success for the digital humanities will have to consist of
connecting the agribusiness of it with local farming and organic gardening.

 
To which she replied: ³Lovely! That's a blog entry if I ever saw one... ³
I quote from this exchange at such length because it may be a useful way of
gesturing towards a space of institutional culture and feeling that is good
to keep in mind while planning for the future.
Helping scholarly end users to decode the encoded
Back to the business of helping scholarly end users to decode the encoded. I
could imagine the TEI taking a much more active and formal interest in tool
development that focuses on getting out what encoders put in. There is
actually a lot of tool building going on. How all the activities are
connected remains an open question. As I understand it, King's College is
about to release a document publishing system that abstracts a common
template from their various projects. Excellent work with eXist has been
done at Brown and Victoria (Canada). Nebraska is dipping several toes into
eXist. The BaseX people have expressed an interest in the TEI. At the US
State Department Joe Wicentowski has been building a site that publishes the
Foreign Relations of the United States with a conceptually simple model that
relies solely on eXist and xquery. Julia Flanders and Scott Hamlin have
recently got NEH grants that fall broadly into the category of outreach and
evangelizing. Then there is TextGrid, which does a lot but could use a
friendlier interface.
 
I am not saying that the TEI should or could go directly into the business
of tool building. I could, however, envisage a meta-role as shaper of a
continuing, active, and structured conversation about goals and principles
of good design, always focusing on the question: "How can scholarly end
users get out the stuff that encoders put in?" I can see circumstances in
which the Mellon Foundation could be interested in funding such a
conversation through workshops of one kind or another. In a more speculative
mode (in a cognitive and economic sense) I could envisage the TEI becoming a
consultant and broker that would help member institutions hire each other's
programmers for specific projects. Angie's List for TEI.
 
You might say that what I am arguing for is already happening at sufficient
scale and that there is no virtue in organizing it further. That is a
serious argument, but I have a hunch that there are quite a few dots that
need connecting and that the TEI can play a useful role in connecting them.
Can the TEI play a role in the collaborative curation of manuscripts?
Returning for a moment to encoding and specifically to encoding of
manuscripts, there seems to be a lot of interest in the collaborative
transcription of all manner of manuscript materials, whether medieval stuff,
Venetian state documents, or Civil War letters. In some of these projects
palaeographical niceties are critical, while in others you want a simple
diplomatic transcription that produces machine actionable texts. I don't
know a whole about this, and I also don't know how much of a role the TEI
plays in such simple and pragmatic projects. The discussions of manuscript
matters on the TEI list tend towards what Michael Witmore calls "the
philologically exquisite." It would be a good thing for the TEI if encoders
at all levels, from the humble to the most learned, would think of the TEI
as the default choice for manuscript transcription. Are we the first name
that comes to the minds of folks in local and state historical societies
when they have a transcription project? Can we lower the entry barriers to
TEI based manuscript encoding?
The hybrid role of the TEI as a standards body, scholarly society and
membership association
I now turn to financial and institutional questions.  The TEI is an odd
body, a hybrid of a standards body, a scholarly society, and an
institutional membership association.  As a standards body it belongs in the
world of acronyms like EAD, METS, DocBook, etc. At least that is where you
find it when you want to start a new file in oXygen and confront your
choices of a document type. A standards body has a single purpose: to
develop, maintain, and improve a particular standard. A scholarly society is
more like a chat room.  It has no single purpose or product but provides a
space of communication, which ranges from gossip to scholarly papers. Its
outcomes are meetings, proceedings, and journals. A membership organization
is more like a lobby: its members justify their support in terms of
particular goals that they want to happen or keep from happening.
 
The TEI-C is a delicate and somewhat awkward balancing act between those
three things. Let me describe the stress points in this balance in terms of
the threats to each of the three identities of the TEI-C. To begin with the
TEI-C as a scholarly society, there is not much of a threat because there is
very little to threaten. I base my observation on financial records in a
Quicken database, which may not record all transaction accurately, but is
probably not seriously misleading.  Between 2005 and 2010 there have been in
any year between 17 and 25 paid-up individual subscribers.  There is not
much continuity and there is no trend line. Most of the members of the Board
or Council have never been individual subscribers (myself included), and
hardly any have been steady subscribers over the years.
 
There is a perfectly innocent explanation for this. Many of the Board and
Council members come from universities that pay institutional membership
fees, and the hardworking members of both bodies may feel that their in-kind
contributions of time and energy are more than enough.  I completely accept
this argument, but it is worth pointing out that the sizable and quite real
international ³TEI community² does not appear to feel a need to express its
sense of belonging through subscription to a society.  From a financial
perspective, the revenue from individual subscriptions is a trivial part of
the budget. I am agnostic on the question whether a membership drive would
be worth the effort.  I am equally agnostic on the question whether teaming
up with the ADHO would make a substantial difference from a financial
perspective. 
 
>From a financial perspective all the Digital Humanities societies are
shoestring operations when compared with other scholarly societies. ASECS,
the society for 18th century studies, has a budget of about $300,000, and
much of its income comes from individual subscriptions. The Renaissance
Society of America is much richer. Its annual budget is about a million
dollars, and it has non-trivial endowment funds.
 
As a standards body the TEI does not face a threat from any competitor. If
there is a threat it comes from people who believe that structural text
encoding simply is not worth the effort. This is a widely shared view in the
NLP and information retrieval community, which is an important source of
text analysis tools and protocols. As I have said above, the TEI needs to
hold its own in that argument and cannot afford to take the utility of its
encoding standard for granted.
Membership association, research libraries, and the TEI budget
That brings us to the TEI as a membership organization, and this is a good
point to produce some budget figures.  By the end of this year the TEI is
likely to have a cash balance of not quite $150,000, which is a good
testimony to a prudent board and its officers. Grosso modo income and 
expenditures have been around $100K+/-10K. We are, I believe, rather better 
off than other DH organizations, but we also face a threat to the main 
source of our income because our major supporters, large research libraries 
are likely to take an increasingly hard look at the value proposition of 
their substantial membership fees.
 
In 2010, the TEI-C collected $92,000 in membership fees from 70 member 
institutions. $60,000 or two thirds of that revenue came from twelve 
institutions paying membership fees of $5,000. Nine of them are American 
research libraries, three of them are institutions in Canada, England, and 
France.  In the current year, two American libraries have canceled their 
memberships, and a third appears likely to do so. In France, we have lost a 
partner but gained a new one. We will lose the Canadian partner next year, 
but may gain a partner in Germany. If you look at the number of $5,000 
memberships between 2005 and 2012, there is, alas, a trend line in the 
numbers 14, 12, 15, 12, 9, 9.
 
Conversations with some librarians have confirmed my sense of what is going 
on. There was a time not so long ago when a chief librarian would readily 
authorize a membership of $5,000 if the cause was generally worthy and some 
faculty member took an interest in it even if the library was not especially 
engaged. $5,000 may have been chump change then, but it is real money now 
and likely to stay that way. No librarian will authorize an expenditure of 
that size unless it can be justified in terms of specific institutional 
priorities. 
 
I had an interesting conversation about this with Merrilee Proffit, whom I 
met a decade ago at the Pisa meeting when she was part of the medieval 
manuscript group.  She is now a senior program officer of the OCLC Research 
Library Partnership, which is the successor of the RLG.  She told me that 
the RLG used to charge membership fees of $25,000 but has now gone to a 
sliding scale between $1,000.00 and $10,0000. From this I gather that the 
TEI-C needs to move towards a more distributed membership model with more 
members and lower fees.  But even with lower fees the ³worthy cause² 
argument is likely to fall on deaf ears.
Asymmetric funding between North America and Europe
There is another asymmetry in the current funding stream, as is apparent 
from this table, which lists the number of memberships in each country, the 
revenue from each country, and the revenue as a percentage of total 
membership income:
 
 
  country   members   total    percentage   
  US   38   $58,750.00   63.8   
  Canada   9   $8,500.00   9.2   
  France   6   $7,441.00   8.1   
  UK   15   $7,425.00   8.1   
  Germany   7   $2,468.24   2.7   
  Ireland   2   $1,750.00   1.9   
  Norway   5   $1,435.00   1.6   
  Belgium   1   $500.00   0.5   
  Denmark   1   $500.00   0.5   
  Netherlands   2   $500.00   0.5   
  New Zealand   1   $500.00   0.5   
  Taiwan   1   $500.00   0.5   
  Slovenia   1   $479.00   0.5   
  Austria   1   $473.00   0.5   
  Czech Rep.   2   $334.00   0.4   
  Bulgaria   2   $266.00   0.3   
  Hungary   1   $250.00   0.3  
 
 
2010 is a typical year in terms of the geographical distribution of funding.  
In looking at this table it is hard not to feel that the TEI-C would be 
better off if roughly half of its income came from North America and the 
other half from Europe.
 
In thinking about how to raise more money from European countries one must 
realize that the American way of raising money through memberships does not 
work very well in Europe. There is more ³top-slicing² of funds, and there 
are no categories in the budgets of library or institute directors that 
memberships can legitimately be charged to. That was made very clear to me 
in recent conversations with Gerhard Lauer and Fotis Jannidis in Göttingen. 
It was confirmed by Merrilee  Proffit who said that her organization very 
much wanted to straddle the Atlantic and that the Europeans wanted to do 
this too but it was difficult to find a funding model.  This seems like a 
silly problem, but it appears to be genuine. 
 
On the revenue side, then, the TEI-C could move along quite comfortably at 
its current rate of operations with a budget of around $100K, where $50K 
comes from North America and its equivalent in euros and pounds comes from 
Europe.  The North American funds would be raised through a more distributed 
model of more memberships at lower cost, while the European funds would be 
raised in some other fashion. It is not, after all, a whole lot of money. 
What about spending?
On the spending side there are a number of uncertainties, and I am not sure 
I understand them fully. Until and through 2010, much of the technical and 
administrative work of the TEI-C was done at four host institutions that 
provided both cash and in-kind services but also were paid for additional 
services.  This system was phased out in last year¹s re-organization. There 
are no anchor institutions anymore. Instead the TEI-C has a looser system of 
³partners² (in principle unlimited in number) who contribute cash and 
in-kind services but receive no payment. The budget will commit funds for 
paid services, but these may change from year to year and are not in advance 
tied to any individual or institution.
 
In the current year, if I read the budget correctly, nobody gets paid for 
anything, but some accounting services will be farmed out. How this will 
work out in practice remains to be seen. 
Is the TEI-C too top-heavy?
I am now going to say things that people may not like, but I am going to say 
them anyhow. That is the outsider¹s privilege, and if you recognize my 
remarks as based on incorrigible ignorance you can laugh me out of court.  
Coming from the outside, looking at the organization charts of various 
roughly comparable organizations, and considering the scope of operations 
and size of membership, I am inclined to think that the TEI-C is a top-heavy 
organization. I am reluctant to say this, particularly in view of the fact 
that there was a substantial reorganization last year.  But how do you 
explain to prospective members why an organization with some 70 
institutional memberships requires a council of twelve and a board of eight 
elected members, not counting non-voting Board members, whether 
representatives of partner institutions or appointed officers like myself?
 
How can an organization with 70 members justify an overhead of two dozen 
people who by virtue of their office can lay a claim for reimbursement for 
their travel expenses, which add up very quickly and are in fact the largest 
item in our budget? I would have a very hard time explaining this mismatch 
to my chief librarian if I tried to persuade her to commit to a full 
membership. So I think we need to give some serious thought to making the 
organization leaner.  There are different ways of doing this without 
damaging the work. One could in fact make a plausible argument that a leaner 
structure would be nimbler and more effective.  Below I sketch a fairly 
radical model that would, however, be compatible with the articles of 
incorporation. Think of it as a straw man proposal to shoot at while 
thinking about the problems that made me suggest it in the first place:
 
1.   The current board and council are replaced by a board of 12 directors, 
at least seven of them elected and the others elected or appointed as the 
membership sees fit. 
2.   The board of directors divides into a technical committee of seven 
members and a general committee of five members.
3.   The chair of the general committee is also the chair of the board.
4.   All offices in the organization are held by current members of the 
board.
5.   The entire board meets once a year at the members¹ meeting. The 
technical committee meets separately on at least one other occasion.
 
There would be an expectation that directors of the TEI-C would have their 
travel expenses borne by their home institutions wholly or in part. This 
³pay for honor² model has much going for it and works equally well on both 
sides of the Atlantic. I was told by Fotis Jannidis that this would be much 
easier to fund in Europe than a membership model. It also is psychologically 
more attractive. Instead of paying a membership fee to an abstract 
organization, a department chair or library head pays for a colleague¹s 
expenses as a form of recognition or career development. There will 
certainly be cases where a potential director comes from an institution that 
simply cannot pay very much, but there are ready ways of dealing with this. 
 
Would a technical committee of seven do a worse job than a committee of 
twelve? Not necessarily. Much of the intellectual work of the TEI happens on 
and off the general listserv. People cycle on and off the council, but in 
their off-years they are still engaged in the intellectual life of the 
community. Even now the work of the council consists less of generating new 
proposals than of reviewing, filtering, and editing proposals that come in 
from the field. Seven people are not necessarily worse at this than a dozen 
and may in fact do better. Besides, the five other directors are likely to 
have strong and useful views about technical matters. So I would not see 
this straw man proposal as involving a reduction in the intellectual capital 
of the current council. 
 
You can see that I am a ³less is more² person except when it comes to 
writing long memos Whether or not this straw man has real legs to stand he 
should make us ask whether we are a little top-heavy and need to think about 
ways of becoming leaner, not simply as a matter of cutting costs but also in 
terms of becoming more effective.
 
Dr. Johnson said of Paradise Lost that nobody wished it to be any longer. 
You may feel the same about this letter. So I stop here and hope that my 
rambling thoughts will at least spur a useful discussion. 
 
Sincerely
 
Martin Muelle