\newcommand{\Lcs}[1]{{\ttfamily\char'134#1}}
\providecommand{\PS}{PostScript}
\def\MP{MetaPost}
\providecommand{\PDF}{PDF}
\title{TUG95 at St. Petersburg Beach, Florida}
\author[Michel Goossens]{Michel Goossens\\\emph{Email: }
  \texttt{m.goossens@cern.ch}} 
\begin{Article}
[\emph{Editor's note:} Michel has kindly allowed us to print portions
of his longer report on TUG 95, which will be distributed on the
Internet.]

\section{Monday July 24th}

Monday morning saw the official opening of the Conference, with TUG95
Organizing Committee Chairperson Mimi Burbank welcoming all
participants before passing the floor on to Michel Goossens, who
thanked everybody for coming, and emphasized the different
structure of this conference as compared to previous years, since this
time we had more workshop oriented sessions (in the afternoons) and
were opening up the area of presentations to the world of electronic
publishing at large, and SGML/HTML, Adobe Acrobat, CD-ROMs, hypertext,
etc.\ in particular.  At that point the representatives of
NTG said a few words about their third edition of the 4All\TeX{}
CD-ROM, and Wietse Dol took this occasion to offer the first pressed
CD to Donald Knuth who was a guest of honour at this 16th
($2^{2^{2}}$) \TeX{} Annual Meeting.  

The first talk was by Ji\v{r}\'{\i} Zlatu\v{s}ka, who showed how \MF{} and
\TeX{} can work together to typeset combinations of text and graphics.
His new approach based on \TeX's extended ligature mechanism, reduces
the number of \MF{} passes  to one, and also simplifies the
\TeX-\MF{} interface.  This permits easier typesetting of text
along curves and in particular allows one to generate beautiful
institutional seals and logos in various forms and combinations
starting from the same base elements.  He noted that, although \PS{}
is often the first choice for including graphics information in \TeX{}
documents, \MF{} often offers improved legibility of the
logos and the letters at smaller sizes.

The next speaker, Richard Kinch, discussed his work on building
reliable \PS{} Type1 and Truetype outlines for the Computer Modern
fonts. He emphasized that several \MF{} primitives (stroked pens,
overlapping ink) have no equivalent in these formats, since they
support only non-overlapping B\'ezier curves.  His program,
\texttt{MetaFog}, handles most of the difficult problems associated
with these conversions. He went on to discuss how \texttt{MetaFog} works,
pointing out some of its drawbacks and making a plea for including
hints inside the \MF{} sources. He finally compared his results to
other outline instances of the CM fonts.

After refreshments Alan Hoenig gave another one of his almost
``perfect'' pedagogical talks, this time showing how one can use
Adobe's Poetica font set, comprising 21 fonts in two families,
exploiting the possibilities of the virtual font mechanism.  Alan
showed us how his macro package together with the font metrics
generated by Alan Jeffrey's \texttt{fontinst} package are able to
typeset a sonnet by Shakespeare as though it were written in the most
beautiful calligraphy of a scribe.

One can never exaggerate the importance of documentation, and to teach
first-year students the benefits of that approach an experiment was
started at Texas' A\&M University by teaching Knuth's \texttt{WEB}
system, that fully exploits combining code and documentation in the
same source. It was found that students who had to ``mix'' program
description and code acquired increased problem solving skills. They
tended to analyse their problems not merely in function of the
programming language used, but in terms of the more general literate
programming paradigm.  Thanks to this increased awareness the students
who took the \texttt{WEB} course also were more successful in grasping
data structures and program development in general.

The morning session was concluded by W{\l}odek Bzyl who showed how by
extending Nowman Ramsey's generic literate tool \texttt{noweb} with a
few stand-alone front-end programs, it became relatively easy to
create a \texttt{\TeX-WEB} system that is easy to understand for the
novice user. The system is extensible by allowing customized styles
and additional features. As an example he showed the literate source
of \texttt{plain.tex}, the \TeX{} source of Knuth's \texttt{plain}
format, and proudly handed a printed copy to Donald Knuth, who browsed
through it with great interest.

After lunch Sebastian Rahtz discussed some advanced exotic features
of Timothy van Zandt's \texttt{PSTricks} package, that provides an
easy interface between the PostScript and \TeX{} languages
(\texttt{PSTricks}'s operating principles were described at
TUG94---see \TUB\ 15(3), pp.\ 239--246).  The talk was a shortened
version of a three-hour presentation by Denis Girou, a well-known
\texttt{PSTricks} guru, at the GUTenberg meeting on June 1st in
Montpellier (France).  By using the electronic version of the slides,
Sebastian could easily zoom in on fractions of text and drawings.
Fractals, complicated curves, like cycloids (this is the first time
that curves of this type published in books on calculus will
actually be drawn correctly, Don Knuth remarked when seeing the
beautiful and precise graphs), and three dimensional multi-colour
calendars were only of few of the graphics gems possible by this
approach.

Jerry Marsden introduced his \texttt{FAS\TeX} system, a library of
standardized system-independent shortcuts for \TeX{} commands.  At
present versions for Mac and Unix exist. This approach speeds up the
keying of input material and greatly increases the accuracy of the
final text. Abbreviations exist for most of the well know formats
and extensions, like the AMS packages. Commands can easily be edited
or added. This approach is especially interesting in an environment
where many non-specialist typists have to work together so that
consistency and ease of input are important considerations.

After tea Denis Kletzing showed how he uses his \texttt{multienumerate}
package to handle complicated list structures. This environment
handles narrow numbered list entries by bundling them in multiple
columns. The drawback of the approach is that the user must specify
the actual layout by typing the explicit position of the entry via
different \Lcs{mitem\emph{ijk}} commands, where \texttt{\emph{ijk}}
are column identifiers. With a little more \TeX{} programming a lot of
the positioning can probably be made automatic, but the approach shows
that it is relatively straightforward to extend \LaTeX{} to cope with
moderately simple but useful structures.

Jon Stenerson described the experience gained by using the style files
he developed for use with \emph{Scientific Word} and which he
described last year at TUG94 (\TUB\ 15(3), pp.\ 247--254). He thought
that the basic ideas of his original approach were still all right but
that most of them will have to be rewritten for streamlining and to
better reflect his present thoughts on the subject.

At the end of the day the editors of the journals edited by the
various \TeX{} User Groups gave an overview of the problems
encountered and ways of improving communication and mutual
re-publication of worthwhile material.  A few weeks before the
conference, an electronic discussion list coordinated by Christina
Thiele (TUG and former editor of \emph{TTN}) and Gerard van Nes (NTG, editor
of \emph{MAPS}) had been set up and allowed the editors to exchange valuable
information.  Presentations were made by Sebastian Rahtz (editor of
the \ukt's \BV), Michel Goossens (for Jacques Andr\'e, editor
of GUTenberg's \emph{Cahiers GUT\-en\-berg} and \emph{Lettre
  GUT\-en\-berg}), Christina Thiele (previous editor of TUG's \emph{TTN}),
Luzia Dietsche (editor of \textsc{Dante}'s \emph{Die \TeX{}nische Kom\"odie}),
Gerard van Nes (editor of NTG's \emph{MAPS}), Wietse Dol (editor of
the Euro\TeX95 Proceedings), Barbara Beeton (editor of \TUB), and
W{\l}odek Bzyl (editor of GUST's magazine). From the various talks it
became soon evident that most of the encountered problems were common
to most user groups' publications, with in particular often an extreme
dependence on a single individual so that when s/he is unavailable
the whole production process suffers. It was clear that a production
team of a few individuals is the only way out of this situation to
ensure that issues can be produced at more or less regular intervals.
Other themes were the difficulty of finding authors, and volunteers to
proofread, correct and edit the articles. All in all, editing a
journal is a non-trivial task and involves the dedication and hard
work of a lot of individuals. As Jacques Andr\'e summed it all up:
``An editor is like an organists at Sunday mass: if the music is good,
no one hears it; if it is bad, everyone cries.''  At the end of the
session Gerard distributed \emph{MAPS} awards to Christina Thiele for her
many years as \emph{TTN} editor and to Mimi Burbank for her hard work
organizing the TUG95 Conference.  

\section{Tuesday July 25th}

As Donald Knuth only rarely attends TUG conferences these days
it was a real pleasure to have him with us at TUG95 and we gave him
the floor for the first part of the morning. After answering the
``usual'' first question---``when is volume four of \emph{The Art of
  Computer Programming} coming out?'' (it will be published over several
years with about 200 pages coming out every six months or so)---he
talked at length about \TeX{} and how he would do it in basically the
same way if he were to start over today.  A more detailed account of
Knuth's presentation will be published elseware based on notes taken
by Christina Thiele.

After the break John Hobby, the author of \MP{}, gave a live
demonstration of his program which is now used by Knuth himself to
prepare the graphical material of his books.  \MP{} implements a
picture-drawing language based on Knuth's \MF{}, but it outputs \PS{}
commands instead.  It moreover gives access to all features of \PS{}
and allows easy inclusion of text and graphics (for more details on
\MP{} see the user guide available on CTAN or an introductory article
in the Euro\TeX92 Proceedings, pp.\ 21--36, Prague, Sep.\ 1992).

The last set of presentations were about ``future systems''.  First
Robin Fairbairns gave an introduction to Unicode. He reviewed the
winding road from 6-bit propriety codes for encoding information on
computers, to 7-bit ASCII-like codes, then to 8-bit EBCDIC and ISO
standards 8859-xx, specific language encodings, the 16-bit East-Asian
JIS, GB, KS and Big Five codes, to Unicode and 32-bit ISO 10646 (for a
detailed discussion of questions of encoding and multi-linguism see,
e.g., \emph{Cahiers GUTenberg} 20, pp.1--53). It is to be noted that Unicode
is the internal encoding used by Haralambous and Plaice's $\Omega$
program, a 16-bit extension of \TeX{} (\TUB\ 15(3), pp.\ 320--324 and
344--352).  Ji\v{r}\'{\i} Zlatu\v{s}ka then brought us up-to-date on
the \etex{} project, whose first version had just been distributed to
developers. Peter Breitenlohner complemented Ji\v{r}\'{\i}'s talk with
some more technical details on \etex{}.  It all sounds like an
interesting development. 

During the first part of the afternoon Wietse Dol showed how easy it
is to use the ``plug-and-play'' 4all\TeX{} CD-system for PC's.  An
interactive live presentation showed the ease with which the system
can be installed from the CD, and also how applications can be run.

\section{Wednesday July 26th}

Mark Swift was the first speaker of the day and in his talk
\emph{Modularity in \LaTeX} he explained that \LaTeX{} should be built
in a highly modular way. In particular an abstraction of functional
modules not mapped onto filenames would be an important point, as
shown by recent discussions on the \texttt{LATEX3} discussion list
where the ``forced'' uniqueness of filenames in the basic \LaTeX{}
distribution, like \texttt{article.cls}, was questioned. The speaker
proposed some possible extensions to \TeX{} and discussed his work on
the \texttt{frankenstein} package, which adds certain kinds of
modularity to \LaTeX.

Bart Wage of Elsevier Science in Amsterdam gave an interesting
description of how journals are handled from source copy to
printed/electronic document. Text is converted into SGML, figures are
kept in various formats (e.g., TIFF, JPEG) and \LaTeX{} sources are
also translated into SGML using the Elsevier DTD.  \LaTeX{} allows for
easy typesetting, but it has no formal DTD, making extensive tagging
somewhat difficult, while SGML allows for a formal DTD, where explicit
tagging of all document elements with respect to that DTD is possible.
Formal specifications exist for math, tables, bibliographic material,
etc.. Elsevier see SGML as an ideal exchange format between different
source representations.  All documents are translated into SGML and
stored in the ``Warehouse'', which forms the common repository for the
various further uses of the documents.  This is extremely important
for electronic documents, where re-use and structure-awareness are of
prime importance. Bart emphasized that a journal is not just a
collection of articles, but a real web of cross-links to related
topics and references, and the future of publishing lies in the
optimization of these facilities for all potential users.

Pierre Mackay then read a paper on \emph{Modern Catalan Typographical
  Conventions} written by Gabriel Valiente Feruglio, who could not
attend. It was an interesting journey in search of typographic rules
for scientific Catalan texts. The author complained that no normative
typographical conventions existed for his language and then went on to
propose a set based on his study of several reference texts. Finally
he introduced a set of possible \TeX{} definitions implementing these
rules.

After coffee Petr Sojka gave one of the best technical papers of 
TUG95---Petr got the ``Knuth'' prize of the Conference for discussing
something important that Knuth did not think about when developing
\TeX---a follow-up of his detailed paper describing the problems
of hyphenation with \TeX{} presented at in Gdansk in September 1994
(Euro\TeX94 Proceedings pp.\ 59--68). This time he discussed the
problems of hyphenating long compound words, which occur very often in
German, Dutch, and the Slavic languages, since in these languages the
constituent parts are not signalled by a hyphen or other fill
character. This makes it often difficult, if not impossible, to
hyphenate words correctly. Therefore Petr suggests extensions to the
hyphenation algorithms of \TeX{} to successfully treat such cases
and he discussed in a generic way which basic functionalities
would be needed. Perhaps something to be implemented in (one of) the
``successor(s)-to-\TeX'', he commented (and Don seemed to agree).

The last talk of the morning was by Sebastian Rahtz, who discussed
the translation of \LaTeX{} sources into SGML. His
presentation was a complement to Bart Wage's earlier that
morning. After working on the conversion problem for several months,
Sebastian came to the conclusion that the only foolproof way is to use
\TeX{} itself to output SGML, a solution implemented by ICPC in Dublin.
He is actually using an intermediate approach, (pioneered by
Sebastian and myself at CERN), which
translates most \LaTeX{} commands into SGML by redefinition of macros,
and then extracts the text from the dvi file. This system 
copes with almost all \LaTeX{} commands (including math).

The afternoon had presentations by convenors of various working
groups. First Norman Walsh presented the work of the \texttt{tds}
(\TeX{} Directory Structure) working group. He explained the rationale
of the choices that have been made, emphasizing that one of the basic
constraints had been ISO-9660, which (only) allows for directories
eight deep and limited to ``8+3'' case-insensitive names for files
(for DOS users this will sound as a blessing, I am sure). Since not
all \TeX{} engines support an optimized recursive directory search
major attention was paid to propose an efficient structure that
minimizes losses of efficiency while searching for package and
font-related files by \TeX. It was emphasized that a production
run-time directory structure, like \texttt{tds} is different in nature
to an archive, like CTAN, and that the two cannot be married easily.

Tomas Rokicki and Michael Sofka then discussed the work in the
dvi-standard committee, especially the standardization of the various
\Lcs{special} commands, that had been discussed by an extremely active
interested group of implementors, meeting over several of ``working
breakfasts and lunches''. I think that real progress was made in this
area where a normative syntax had been awaited for too long. I am
very grateful for the enthusiasm shown by these people, and am
convinced that we shall see their work bear fruit in the near future
(Tom told he will be working actively on his program \texttt{dvips}
over the next few months, so that we can be sure, knowing Tom's
reputation, that many of the hyper- and other goodies discussed during
the conference, will become part of this, and other popular
dvi-drivers).

After the refreshment break T.~V.~Raman gave a practical demonstration
of his ASTER system, which allows one to ``hear'' \LaTeX{} sources,
including mathematical formulas, being read out. His system uses a
speech synthesizer via an augmented emacs editor running with Common
Lisp, and is  able to analyse, decode and then transcribe into
audible form well-structured \LaTeX{} documents.  This last remark is
extremely important, since, as already pointed out by Sebastian Rahtz
in his talk earlier that day, due to the various (ambiguous) ways that
mathematics can be coded in \TeX{}, there exists no automatic way to
parse such \TeX{} source into something usable more generally, such as
SGML or audible sound. This was the loudest plea yet for using
well-structured markup.

Just preceding the Conference dinner the editors of the various
\TeX-related magazines had a second meeting to discuss ways of
improving communication. It was decided to write a short overview of
the experiences of each team for \emph{TTN}, to provide cross-references to
each other's publications on the user groups' respective WWW pages,
where tables of contents of the magazines will be posted (in fact,
GUT\-enberg already decided to make freely available on the Internet
via WWW, all articles---initially in \PS{} form only---of the
\emph{Cahiers} and \emph{Lettre GUTenberg}). It was also proposed that
all non-English publications try and provide an abstract in both the
local language and in English, so that these abstracts can be
published in \TUB\ (or elsewhere).  Editors were also asked to
signal potential articles that might be interesting for translation
into English and publication into \TUB\ (of course, editors can
translate articles from \TUB\ into their national language also!).
Presently, CSTUG, GUST, GUTenberg, NTG, TUG, and \ukt\ are working on
a TEXART CD-ROM that will make all publications of those user groups
available on this electronic medium (and on the Web).  It was also
agreed that each author should be asked permission to reprint her/his
article(s) in this way.  During the meeting Gerard had the pleasure to
offer the third \emph{MAPS} award to Barbara Beeton, the most senior and
long-standing editor present for her 16 years of efforts to make
\TUB\ an example of the typographic quality that can be achieved
with \TeX{}.

At the end of the conference dinner in the evening a set of books
written (and dedicated) by Donald Knuth were put up for sale to the
highest bidder. Even Lamport's \LaTeX{} book got a nice inscription by
the hand of Knuth.  After I let the the volumes of the ``Art of
Computer Programming'' escape I concentrated on the ``big one'',
namely the five volumes of \emph{Computers \& Typesetting}, who became
mine for the nice little round sum of \$700. Adorned with the
dedication of Donald Knuth this will remain one of the treasures of my
personal library!  The sales were an outstanding success and about
\$1800 were collected towards the funding of the Euro\TeX{} bus which
will take participants from Russia and central Europe to the
Euro\TeX{}95 Conference in Papendaal (the Netherlands).  Many thanks
to Addison-Wesley, who donated the books.

\section{Thursday July 27th}

In his presentation T.~V.~Raman gave an overview of ASTER---an Audio
System For Technical Rea\-dings---the system he demonstrated the day
before. ASTER renders \LaTeX{} documents in an audible way, so that
visually impaired persons can ``listen'' to their contents.  Raman
emphasized the importance of the use of clear generic markup for the
input source document to ease the extraction of structural logical
information that can be easily translated into an internal
representation. ASTER then renders information by applying rendering
rules written in AFL---Audio Formatting Language---to the internal
representation. In a sense AFL is to audio formatting what \PS{} is to
visual formatting (although AFL is by far not as complex).  As a
conclusion he emphasized that one needs a semantic-oriented DTD to
produce a high-quality audio document. Since no such completely
general DTD can be constructed, one has to use the facilities provided
by \LaTeX{} and its hyper\TeX{}t extensions.

Mark Doyle next reviewed the purpose and history of the Los Alamos preprint
server, which is one of the first (and more successful) document
servers on the Web. In fact it started in the area of
(theoretical) High Energy Physics and took place in close
collaboration with CERN (where WWW was ``invented''). Today several
tens of thousands of preprints are available online and over 20000
users visit the server each day. Although at present most documents
are only available as (mainly \TeX) source and standard \PS{}, they are
now producing \PDF{} versions that include cross-references to
other documents on the Web using the \texttt{hyper\TeX} tool and \PDF\
hypertext links. In this way cross-references to other
documents can be easily instantiated.

During the next half-hour I gave an introduction to Nikos Drakos' tool
\texttt{latex2html} and showed how by simple customization the visual
quality of the output HTML files can be substantially improved. I went
on the show how the \texttt{latex2html} system also allows for
interconnecting separate documents. I ended with a few examples of
HTML3 output generated by an ad-hoc program developed at CERN and
viewed with the HTML3-capable \texttt{arena} browser.

After the break Sebastian Rahtz showed how with his \texttt{hypertex}
package (sharing some code with the \texttt{Hyper\TeX} package
discussed earlier by Mark Doyle) it is easy to turn \LaTeX{} documents
into hyper-documents. Their ``hyper'' contents can be enriched by
adding supplementary information about \LaTeX's cross-references via
\Lcs{special} commands.  These are picked up and translated into
\PDF{}'s \texttt{pdfmark} commands by Mark Doyle's ``hypertext''
\texttt{dvihps} program, an extension to Tom Rokicki's \texttt{dvips}
program.  Tom stated that these extensions will end up, in one form
or another, in the forthcoming upgrade of standard \texttt{dvips}.

The afternoon started by a second presentation of the \emph{4All\TeX}
system, and, as always, there was great admiration amongst all those
present for the ease with which it is possible to ``plug and play'',
i.e., start to setup and run the system without much ado.  It became
all the more evident that such a CD-ROM for Unix is a real need, and a
recurring proposal for the next great thing that TUG should come up
with (and we are surely thinking about a way get this done).

During the next hour I gave an introduction to SGML using HTML as an
example of a DTD, and showed that it is not difficult to understand
the structure and syntax of a DTD, and from there to figure out the
various possible document elements, their attributes and the entities
that are available to the user. Work on other DTD's for mathematics
and tables were briefly mentioned, as were a few tools for authoring
and checking SGML documents. I came away with the feeling that at the
end of my talk most of the audience had a more balanced view about
what SGML is, and what it is not. I therefore hope that my presentation
will also contribute to eliminate most of the artificial animosity
between the SGML and \TeX{} worlds. As Sebastian, myself and a few of
the other speakers tried to show, SGML is about structure, and \TeX{}
about typesetting, and the two tools are therefore complementary and
both useful.

Chris Hamlin, in the last scheduled talk of the day, described the
production work at the American Physical Society, and, as expected, it
is similar in content, form, and structure to what we had heard by
other speakers (at this conference Elsevier, or at other conferences
Springer, OUP, etc.), namely a mixture of \TeX{} and other
word-processor inputs are accepted by the production team.  The
proportion of \TeX{} sources varies wildly between publications
(between almost nothing in the chemical journals to well over 50--60\%
in some of the physical journals). Various house styles are available,
and at present ways are being investigated to translate the inputs
into SGML to take full advantage of electronic publishing tools.

The last part of the afternoon was for the TUG Business meeting.

\section{Friday July 28th 1995}

Already Friday. It seemed as though the Conference only just started,
but the bags at the sides of the room and the now-empty vendor's room
made us realize that we were only here for another few hours.

The morning started with a paper submitted by Jonathan Fine, but read
by Alan Hoenig in Jonathan's absence. The title was \emph{New
  perspectives in \TeX{} Macros} and dealt with a possible way of
combining the advantages of both SGML and \TeX. His \TeX{} macro
package \textsc{simsim} takes SGML and style files as input and
generates pages formatted with \TeX{} as output. \textsc{simsim} comes
with an SGML parser and the style files are used to link \TeX{}
actions to SGML events. The \textsc{simsim} system also offers a
programming environment for writing \TeX{} macros and style files.  At
present issues of performance were not addressed directly but on
sample documents the speed was comparable to that of \LaTeX{}.  All by
all an interesting idea, and I look forward to see Jonathan's finished
product soon.

Sergey Lesenko then told us about his \emph{t1part} tool that
partially embeds Type1 \PS{} font files into a document.  The
principle is to include the \PS{} code for only those characters that
are actually referenced. This can result in huge savings in size if
one uses only a few characters from many fonts (the procedure is based
on the same model that includes only the necessary Type3 bitmaps for
characters built with \MF).  Tom Rokicki and Sergey have been working
together over the last few months and this facility will be built into
the ``next'' version of Tom's \texttt{dvips}.  I mentioned that Basil
Malushev has a somewhat similar utility \texttt{fload}, that uses the
publicly available \texttt{ghostscript} program to make a map of all
referenced fonts and then includes only the characters needed.
Basil's approach can be used for any kind of \PS{} file, so that it is
complementary to Sergey's which is well integrated with \TeX{} and
needs no supplementary external program.  During the discussion there
were some interesting remarks on copyright issues connected with
including Type1 fonts inside documents.  It was felt that, although
partial font loading would make pirating fonts less effective, it
does not mean that all font vendors would agree to let us include
their fonts in this way in files distributed electronically (on CD-ROMs
or the Internet). To be continued\ldots

A more technical talk, on \MF{} this time, was Jeremy Gibbons'
presentation \emph{Dotted and dashed lines in \MF}. He showed that
drawing evenly spaced dotted and dashed lines in \MF{} is a
non-trivial task, and he proposed several solutions to make it
possible. He introduced the notions \emph{evenly spaced in time} as
opposed to \emph{equally spaced in space} and went on to show that
they are far from identical, since points can move at different
``speeds'' in space as they progress along a path evenly in time.
Using recursive adaptive refinement techniques he showed how one can
solve the problem in \MF. His procedure can be extended to allow for
dashed, or alternating dashes and dots. As recursive techniques have
the unwanted feature that they can overflow the stack, Jeremy also
proposed a solution based on an iterative non-adaptive technique that,
although perhaps less elegant and automatic, does the job almost
equally well.  At the end of his talk he showed several attempts at
drawing an attractive muskrat, the logo of the \emph{Mississippi
  Muskrats} jazz band he used to play in.

The last scheduled talk was by Robin Fairbairns.  After explaining the
principles of the \PS{} multiple master Type1 font format, Robin
showed how a crude first system of using these fonts with \TeX{} was
set up.  All font instances are expressed in function of weights with
respect to the master designs. These weights are calculated by the
\PS{} interpreter from the design parameters via the \PS{} operator
\texttt{ConvertDesignVector}. One has to use version 3.x of the
\texttt{ghostscript} program to extract the weights, which were then
used to generate the Adobe Font Metrics (AFM) instances from the AFM
files for the master designs.  Then Rokicki's \texttt{afm2tfm} program
was run to generate corresponding tfm files needed by \TeX{}, while a
header file was also defined to allow \texttt{dvips} to actually
specify the font instances from the weightvector. This set up was used
to typeset the last issue of \ukt's magazine \BV\ in
Minion, one of Adobe's Multiple Master fonts.

The morning ended with the ``Closing Ceremony'' and the announcements
of the TUG95 prize winners.

Christina Thiele, as vendor and public-relations liaison thanked the
various companies who had vendor booths or otherwise contributed to
the TUG95 conference, in particular Addison-Wesley for the books they
donated (and that were put on for sale on the Wednesday evening for
the Euro\TeX95 bus, and still a few left for another sales idea we are
playing with for spicing the TUG96 bursary. Stay tuned to \TUB\ or
\emph{TTN}!).

Mimi Burbank, as Chair of the Organizing Committee, thanked all the
people at SCRI who helped her financially, organizationally, by
providing PC's or a printer. She also thanked the extremely efficient
hotel staff for their never-ending devotion to a job well-done.

Sebastian Rahtz, the chair of the Programming Committee, then
announced the prize winners for TUG95.  Just before coffee that
morning all participants were asked to write down an ordered list of
the four papers they liked most, and on the basis of that list it was
Raman who was selected as best presentation, best paper, and most
important contribution to the \TeX{} world (and humanity, one person
wrote).  The Knuth prize, for the paper discussing something that he
``forgot'' in his \TeX{} program, went to Petr Sojka for his work on
hyphenation.  Other prizes went to Richard Kinch for \texttt{MetaFog}
(who put his prize copy of Textures up for sale, so that the
Euro\TeX{} bus got again somewhat more money to take home), Alan
Hoenig for his marvellous Poetica work, Jeremy Gibbons for his
entertaining and erudite explanation of \MF, and Sergey Lesenko and
Tom Rokicki for partial font downloading, and work on dvi standards.
Many thanks are due to the fine \TeX\ vendors Blue Sky Research, Y\&Y,
PC\TeX, and Richard Kinch, who generously donated copies of their
products for the prizes.

Of course we did not forget
our friends from NTG, without which this conference would not have
been the same. Their \emph{4All\TeX} CD-ROM was one of the highlights
at this conference (they sold about 40 copies, and the remaining 60
were taken to the TUG office for selling them to the North American
TUG community). Their ``presence'', good humour, the organization of
the book auction, the coordination of the TEX-ED initiative and the
hundred or so photos they took made them a memorable and unforgettable
part of this meeting. Therefore a signed copy of the \MF{} book was
given to Wietse Dol.  In a gesture underlining their dedication to
\TeX{} and TUG Wietse then offered TUG the two golden (original)
CD-masters of the third edition of \emph{4All\TeX} that just appeared.
I had the pleasure to receive them in name of TUG and I promised that
they would be framed and displayed in a prominent place in the TUG
office in San Francisco. The gifts were concluded with the UK \TeX\
Users Group and TUG presented 2 bottles of wine, and 2 boxes of
chocolates, to Don Knuth, maintaining the `2' theme begun by NTG's 2
CDs at the start of the conference.

Finally it was my duty to formally end the TUG95 meeting, and after
thanking Knuth for his presence, which made this $2^{2^{2}}$th meeting
even more special, I re-iterated the thanks to all vendors, SCRI and
the hotel staff, for their display of (southern) American hospitality.
Then I invited all participants to the next (17th) TUG annual meeting
in 1996 in Dubna (Russia, 150 kms north of Moscow, on the Volga
River), where from July 28th to August 1st TUG96 will be hosted by the
Joint Institute of Nuclear Research.

During the afternoon Alan Hoenig gave a partical introduction to the
use of virtual fonts. He showed how they can be used to create new
characters as various combinations of glyphs and rules. He described
how Alan Jeffrey's \texttt{fontinst} package allows one to easily
install \PS{} font families. In his usual pedagogical approach Alan
made it all sound as though it is extremely simple and
straightforward, and all fifty participants to this last ``event'' of
the conference came away with the feeling there were ready to generate
some virtual fonts themselves.

\end{Article}