%%%%% CaesarCM.doc, April 1992; version2, Sept 1992, Jan 1993



                    === CAESAR-CM ===

  **** A SMART HYPHENATION AND ACCENTUATION MACRO PACKAGE ****

               --- documentation ---

                  Laurent Siebenmann

     The name Caesar originated as the "non de guerre" of the Roman
conqueror of Gaul. It is often repeated that "he was untimely ripped
from his mother's womb". Our use (like the Romans') is explained by
the Latin verb caedere, to cut, (past participle caesum), and the
nouns caesura, the cut, and  decisio, juridic decision. The present
use of the name Caesar comes more immediately via the French
c\'esure, meaning a pause within a verse following an accented
sylable, or (more recently) hyphenation.

     The CaesarCM package was built to provide a flexible
transitional solution to the problem of providing good hyphenation
for French and (US) English under TeX implementations that for one
reason or another are equipped with Computer Modern (CM) fonts
rather than with fonts based on the conflicting (but improved) TeX
norm of Cork 1990. Flexible is the keyword --- the hyphenation will
be perfect for those fonts you have space to install in a suitably
extended form, and it will never be worse than the respectable
hyphenation that the HyAcc-CM package provides for standard CM
fonts. It will probably remain in use until all TeX installations
have compact virtual fonts in use which follow the Cork norm AND
provide emulation of the original CM norm.

     Although it was built with French users in mind, this package
is not really specific to any particular language. Admittedly, users
of some nationalities may not yet find all the features they want.
(Germans will however find that ^^bd can serve as compound word mark
<cwm> in the TM fonts specially prepared for CaesarCM.)  Hyphenation
patterns for many European languages are posted on

     ftp rusinfo.rus.uni-stuttgart.de   [129.69.1.12]

     This package may be already out of date for PC microcomputers
thanks to the great strides toward optimal implementation of the
Cork norm using virtual fonts by the TeX implementations of
Brouard-Sullivan and more recently Eberhard Mattes.  MLTeX, an
extension of TeX by M. Ferguson, has always offered excellent
hyphenation.

     In contrast, no firm schedule is known for the implementation
of virtual fonts nor any form of MLTeX with a standard Macintosh
interface (eg under OzTeX and Textures). This statement has to be
clarified: for the unix-like Macintosh Programmer Workshop
environment there is since early 1992 a TeX environment called
DirectTeX by Wilfried Ricken which has rudimentary virtual font
support. (Available by anonymous ftp from ftp
rusinfo.rus.uni-stuttgart.de [129.69.1.12].)

     The major new feature of this macro package is flexible use of
a number of (possibly partially) extended Computer Modern fonts,
to obtain improved or optimal hyphenation for words typeset in
these fonts. The prime goals here are to save disk space and
maintain 100% compatibility with current TeX documentation and with
archives of .tex files.

     For fonts that are not extended at all, one gets, by default,
an intermediate form of hyphenation due to Jacques Desarmenien; it
is known through his French-English Plain format called
feplain.fmt, and it gives good results except near accents. This
intermediate form of hyphenation has been recently updated for
international use by D. Flipo and the author.  If your environment
is "CM" but you do not have any extended CM fonts and are not in a
position to install any, then you should use the smaller package.
(I have travelers and American institutional users in mind here.)
It is called HyAcc-CM.tex and is available along with this package
on ftp rsovax.ups.circe.fr [130.84.128.100].

     As a core sample to start from and build on, you will find
with this package extended versions of the most commonly used prose
fonts, cmr10, cmtt, and cmti10. They are provided in generic TeX
format, and with some additional variants to serve the Macintosh
TeX implementations Textures and OzTeX. These extensions come from
a full collection "TM" of such extensions to the ISOLatin/1 half
192-255 of the Cork accented letter range 128-255  that was
provided in February 1992 by D. Taupin on

     ftp rsovax.ups.circe.fr  [130.84.128.100]

(login: anonymous).  This package is also available there.
Beware however that all .tfm files used for extended CM fonts
to be used (to advantage) by CaesarCM must be suitably marked
either internally or by name. An appendix describes just how.

    The net disk-space cost of the quite ample "TM" extension of
(for example) cmr10 is is under 40K octets with all magsteps
present in .pk format. In contrast, the full .pk version dmr10 for
the Cork norm would amount to about 160K, of which half might be
saved by jettisoning the CM norm version. But more importantly,
without the flexibility this package offers, it would be necessary
to upgrade all prose fonts at once at a cost of several megaoctets.

    Daniel Taupin's aim in posting the "TM"  fonts was to provide a
straightforward continuation of the CM regime that taps the
hyphenation powers of TeX available since fall 1989. This package
arose as a result of my efforts to make efficient use of his fonts.
His original recommendations required all fonts to be loaded under
their "TM" names, and it imposed a trivial, but not completely
hassle-free, revision of all formats and styles. In contrast I
recommend that "TM" fonts be installed under "CM" names, because it
obviates revision of preexisting format files; you just have to
recompile as indicated below.

    The abstract notion of a partial extension of CM fonts,
on which this CaesarCM package is based, is made precise in the
Appendix on EXTENDED CM FONTS. It would be fair to say the case of
TM fonts is a preexisting ISO based standard that the architects
of the Cork norm set out to enlarge and improve.

    Purists may be irritated by my continued use of the name "CM"
for the extended fonts, rather than "TM", or "dm" an earlier term
due to Brouard-Sullivan for this sort of extension (see their TeX
package for PC ibid.).  I disagree: 100% upward compatibility by
simple extension of CM following an undisputed norm justifies
maintaining the name  "CM".  The resulting simplicity would be
reason enough to use the term "CM" even if it were "immoral".  In a
many-user environment those who are happy with the original "CM"
environment need not be aware that anything has changed --- just
as with MLTeX.

     The use of the designation Extended Computer Modern for fonts
of the Cork norm is more deceptive since the Cork norm
significantly disrupts the Computer modern.

     See a full description given below of more useful switches at
the users disposal. It should be emphasized that the average user
will never need or want to make any adjustments whatever.  They are
made available for programmers and typographers.

     Suppose that your TeX environment has space enough for the
full collection of extended computer Modern fonts, say as
distributed by D. Taupin on rsovax.circe.fr.  Is there advantage
to be derived from the use of this CaesarCM package? I suspect the
answer will be yes for a very long time to come.

     Firstly, the Computer Modern style is just one style of many,
and since even the Taupin fonts are not fully extended, it is
eminently reasonable to allow for other fonts extended over a
different subset of the Cork "semi-norm" 128-255.  The fonts being
produced by J. Zlatu\v ska <zlatuska@cspun12.bitnet> give an example
(see Cahiers GUTenberg 10--11, pages 57--68). It is worth
emphasizing that the "smart accents mechanism" of CaesarCM allows this
by operating on a fontclass-by-fontclass and character-by-character
basis. Secondly, the world has quite a population of unextended CM
fonts such as Concrete Computer Modern ("ccr"), Pandora, Lucida. 
CaesarCM gives perhaps the simplest way of using these along side of
nontrivially extended fonts. For example, it might be reasonable to
do the main text of a schoolbook in tenpoint "TM", and the exercises
in ninepoint "ccr" plus Euler math fonts.

     As mounted by Textures on the Macintosh, the PostScript fonts
are, in effect, exotically extended CM. (They have some
miscellaneous chacacters >127 but not in normalized positions.)  By
default CaesarCM treats them as unextended CM fonts.

    Here is a helpful but incomplete description of how Knuth's
accent syntax \', \`, \^, \" etc. acts after this package has been
loaded. A more complete description is provided in appendix xxx.
Consider \'e for example (in prose). The extension class of the current
font \font is normally identified by a large negative value of the
function \skewcharacter\font. Then the presence of a ready-made
"e-acute" character for this class is tested at the location
specified by norm --- by testing whether a macro labelled by the
class and the position is defined. If a ready-made character is
detected, it is used. In other cases the approach of HyAcc-CM is
followed: Knuth's \accent macro is used to form a composite
character, and the trick of Desarmenien is applied to allow hyphens
in the same word but not too close to the accented character.

    For French, nearly all hyphenation breakpoints become available
if just <e-acute> and <e-grave> are available in the extended CM
font.  Such a miserly extension of CM applied to all prose fonts
and then further enhanced by generous extensions of the major prose
fonts by the "TM" series might offer the most economical use of
this package in a typical French microcomputer environment.

    To benefit from this package, accents must be produced via Knuth's
accent syntax \', \`, \^, \" etc., but the use need not be direct. It
is possible to type ready-made accented characters, provided these
characters are made active (macro-like). For example definitions of
the form

   \catcode`\<e-acute>=13 %% active category
   \def <e-acute>{\'e}

will do the job. Most TeX sites offer a suitable file of such
definitions. Avoid definitions of the form

   \edef <e-acute>{\string^^e9}

or 

   \def <e-acute>{\char"E9}

as they will only work when the current font has an <e-acute>
present in the slot "E9 provided by the Cork norm.

    This package wraps up all the considerable complexity of
previous macros for accents and hyphenation, notably those coming
from M. Ferguson's MLTeX. I have had to simplify and compact them to
build something more flexible. This process naturally introduced
some bugs. Some may remain, so please report untoward behavior
promptly!

    I offer sincere thanks to several colleagues for various forms
of assistance.  I have already mentioned recent contributions of
Daniel Taupin and Daniel Flipo that provided vital ingredients.
Laziz Abdelouhab piloted Victor Ostromoukhov's Mac Metafont for me
under MPW, and also the Blue Sky Research font tools.  Barry Smith
and Doug Henderson provided me with clues on fonts for Textures
after generously allowing the posting of their font tools in July
1991.  I hope others yet will go the final lap to install this
mechanism for the multitude of users who lack familiarity with TeX
fonts but need improved accents and hyphenation at very modest cost
in disk space.

    A quick test of this package, is indicated in CaesarTeaser.doc.
For permanent installation, instructions are given below.  

     Most users making a permanent installation of CaesarCM will
want to simultaneously install hyphenation patterns, and other
facilities for peculiarities of the language(s) to be used. Such
installations of several major formats are provided for French &
English by fe-format-dumper-cm (same postings); it can hopefully
serve as a model for dealing with other languages.


  --------------------------


COMMAND SUMMARY

     Most users should be happy with the default settings of
CaesarCM.  Thus the following commands are mostly for programmers; 
their description tells a lot about how CaesarCM works.

 --- \Caesar, \noCaesar :
 These turn on and off the basic smart accentuation mechanism of this
package. A ready-formed character-with-accent is used whenever
available for the current font class, and otherwise an accent is
combined with an unaccented character.

 --- \ShowCaesar, \HideCaesar :
 Diagnostic commands to show on the TeX log when a
character-with-accent is used via the "smart accents" mechanism,
and when a font is identified for the first time as an
extended CM prose font.

 --- \ShowCurrentFontClass 
 Shows on the TeX log the font class Caesar has assigned
to the current prose font for the purposes of accent formation.

 --- \skewchar<font>=\CaesarClass<class>
 manually reassigns the class of a <font> like \tenrm to a class
like CM or TM.  Caesar usually does class assignment automatically on
the basis of internal marking of the .tfm or the font name.  This
command lets Caesar exploit fonts with unmarked .tfm's having 
arbitrary name.

 --- \JDaccents \noJDaccents :
 These turn on and off a Jacques Desarmenien mechanism that allows
hyphenation far from the accents in a word with composite accents.
Roughly speaking a word-break is introduced between the accent and
the underlying letter, but zero space is actually inserted and no
line break is allowed there.

 --- \MLcodes, \noMLcodes :
 Let \ae, \Ae, \oe, \Oe, \o, \O, \i participate (respectively do not
participate) in the hyphenationation process and in action of
\uppercase, \lowercase.

       By default one has \Caesar, \JDaccents and \MLcodes. Provided
no accents and no exotic characters occur in a given word, this
still gives the same possible breakpoints as Knuth's standard Plain
TeX 3.0. If for some reason you need its rather unpleasant behavior
in the presence of accents, use the unpleasant command
\noCaesar\noJDaccents\noMLcodes.

 --- \RawPSText, \CMText :
 Some implementations of TeX, for example OzTeX (but not Textures),
offer no ready-made adaptation of Adobe PostScript fonts to the CM
font encoding.  Then a good many things have to be redefined when
one passes from the default Computer Modern regime to the "raw
PostScript" regime. This is the job of \RawPSText, and \CMText brings
you back. Do not forget to load PostScript fonts (via \font) before
trying to use them. Also, be alert for portability problems,
inevitable when different standards clash.  The raw Postscript font
encoding includes no ready-made accented characters.

 --- \showhyphens{\myfont ...} :
 This is standard Plain TeX, but there is a new twist in its
behavior. It serves to indicate the possible hyphenation points
available in a piece of text .... What is new is a possible
dependance on the (text) font \myfont you use --- and CaesarCM is
responsible for that dependence. Beware that if you do not specify
the font explicitly, Plain TeX uses \tenrm --- NOT the current
font! Note also that the breakpoints shown may be unavailable
because \lefthyphenmin and \righthyphenmin are greater than 1.

     The following commands are for combatting 
occasional bugs; see Caesar.bugs for details.

 ---  \noDSaccents, (and \DSaccents to return)

(DS meaning Double Strike) serves to combat low resolution printer
smudging of composite accented characters.

 ---  \ufcr or equivalently \UpdateFontClassRegister

serves to restore an occasionally suppressed kerning before the first
accented character after a font class change.

  ========================================

     The following two Steps (A) and (B) make CaesarCM fully
functional.


 (A)  FONT AND METRIC UPDATING

     To derive any benefit whatever the novel feature of the
CaesarCM mechanism, you will want to add some CM norm fonts that
have been significantly extended to the range 128-255 in accordance
with the Cork norm of 1990 (see possibilities above), and been
suitably marked as extended.

     If such a font goes under a new name, say tmr10, you can
install or remove it just like any other font, and then go straight
on to step (B).

     For serious use, it is recommended that you install extended
versions of your principal prose fonts, normally cmr10 and cmti10,
and add more if the need arises.

     We consider the case of bitmapped fonts of the .pk norm. Each
font comes in two parts: a .tfm file such as cmr10.tfm containing
all the metric information TeX needs and a collection of .pk files,
one for each magstep. Each .pk file contains the character
bitmapped images for the printer.

     A copy of the extended version of each .pk file should replace
the original; the name should be that of the original, say
cmr10.pk300 not tmr.pk300.

     One does the same with the .tfm file. To create or obtain a
suitable modified .tfm, see an Appendix.

     It is helpful for the user to be able to readily distinguish the
extended CM fonts from the unextended ones. The devices for external
identification vary. Size, date, and name are always helpful. For OzTeX
and Textures on the Macintosh, identification by icons is provided (see
Appendices). Once CaesarCM is loaded identification is easy.
\ShowCurrentFontClass tells the class of the current prose font (the one
identified by \font).  For example (under plain)

     \it \ShowCurrentFontClass

will tell you (on the log) the class of the font for italics.

     Fonts can be added just a few at a time, as needed to maintain good
hyphenation. However, beware that if you extend a standard font that is
used in your precompiled format, then the smart accents mechanism cannot
use the new accented characters until your format is recompiled as in (B)
below to absorb the new information in the corresponding .tfm.

  --------------------------

 (B) COMPILATION OF CaesarCM COMMANDS IN A Plain FORMAT

      It is important to use version 3.xx of TeX, and a version 3.xx
of plain.tex.  The critical date is autumn 1989, and earlier material
will be inadequate.

      Here are instructions for a generic TeX, with Plain Format and
French language and just to be specific. More elaborate formats will be
used in practice and are discussed below.

      fhyphen.cork is (or calls) a copy of the French hyphenation
pattern file using the ISO Latin 1 (or Cork) standard for encoding
accented characters.  Place together in one directory:

 (i)  a copy of fhyphen.cork renamed hyphen.tex

 (ii)  a copy of plain.tex exactly as provided by D. Knuth

 (iii) a file x.tex whose contents is the single line:

  \input plain.tex \input CaesarCM.tex \dump

Note that some systems (Mac, VMS, ...) identify upper and lower case
letters, while others (unix, ...) distinguish them.

     Compile x.tex using INITEX or VIRTEX, i.e., the primitive form
of your TeX made for format building.  After launching this TeX, one
responds to a prompt ** by typing

**x <carriage return>

It does not matter if minor fonts such as manfont are missing; just
hit carriage return. In time you should  be invited to name the
resulting format. Choose a distinctive name; it is often useful to
indicate the TeX implementation and/or configuration used, since a
format will normally run only with its parent TeX binary.

     With this format you can run various hyphenation tests with
various fonts and font configurations.

     Where French is concerned, the compilation of formats can be a
more complicated matter. This is because it is vital to accompany
French hyphenation by French spacing and punctuation through the use
of active characters, and this still causes complications with many
formats such as AmSTeX and LamsTeX. Thus, a special French(-English)
format compilation mechanism has been provided. It is called

            fe-format-dumper

(or fe-fmts for PC's) and is posted as a directory on on ftp
rsovax.ups.circe.fr [130.84.128.100].  It is not part of CeasarCM
which is to be kept basically language independant. Further,
fe-format-dumper serves HyAcc-CM as well as CeasarCM.


WHEN SHOULD YOU RECOMPILE?

     You should recompile your format, whenever you extend a
font it uses. This is because the old .tfm file is already in
the compiled format; thus to take account of the new .tfm you must
recompile!  It is true that this can be a pain if you are 
unprepared; fortunately there numerous exceptions to this rule.

     This need arises from a weakness of TeX: it cannot unload a
font of a given name, and it will never load a .tfm of a given
name xxx a second time (via a command of the form \font\ft=xxx).
Instead it reuses the .tfm named xxx that it first loaded.

     On the other hand, if the fonts in question are defined for
the first time in your style file (not your format),
then there is no need to recompile your format.

     Many users have formats that are not to be recompiled, perhaps
because the the ingredients are not readily
available, perhaps because you dislike format compilation. In such
a case, it may should be possible to place yourself in the above case
mentioned where no recompilation is needed.  Suppose for example
one wants to further extend cmr10, say from a version including
just French accents to one with all european accents. WITHOUT
RECOMPILING ONE CAN INTRODUCE THE EXTENSION UNDER A FRESH NAME, say
eumr10.  To use this font instead of cmr10 it usually suffices to
put

     \font\tenrm=eumr10
     \rm

in your style file.  This should override \font\tenrm=cmr10 in your
compiled format. This is quick and quite practical, but it is best to
regard it as a temporary measure, which if pursued could lead to a
bloated and anarchic font collection.

  --------------------------


                   Appendix  EXTENDED CM FONTS


      By a CM extension class we will mean a subset E of the accented
characters of the range 128--255 of the Cork norm.

      By a (partially) extended CM TeX font "belonging to" class E will be
meant one extending the Computer Modern norm for alphabetical fonts in the
range 0--127, and in agreement with the Cork norm for the subset E of the
range 128--255.

      In practice it is reasonably clear what "extending the Computer
Modern norm" means. It is perfectly clear when, in the range 0-127, your
font coincides in all respects (including name) with one of Knuths CM
fonts. But note that cmr, cmti, and cmtt differ slightly; for example cmti
has a pound sign where cmr has a dollar. The Cork norm of 1990 eliminates
such variations.

      The CM fonts of Knuth, the "TM" fonts of Taupin, and the
virtual fonts of J. Zlatu\v ska are naturally of extended CM type.

      Unfortunately there exist incompatible extensions of "CM"
fonts, notably the numerous adaptations of Adobe Postscript fonts
used by Textures on the Macintosh.  (Many other TeX's use raw
PostScript encoding, which makes these fonts "non-CM" even in the
range 0-127; Caesar does not deal with them at all.)  Such
noncompatible extensions can be regarded by CaesarCM as 
simply CM fonts (ie as extended CM fonts of class E where E is
the empty set). The extra characters do no harm, but they cannot
be exploited by the smart accents mechanism of CaesarCM.

      For an extended CM font to be useful to CaesarCM, an
extension class E to which it belongs must somehow be made known
to CaesarCM, which is then prepared to use the ready-made
accented characters that belong to E.

      Remember then that an extended CM font of class E
may have some (other) accented characters contradicting the 
Cork norm; CaesarCM just ignores the latter.

      Each new class must be  made known to CaesarCM.tex by
adding suitable data.  Currently, just three classes are known to
CaesarCM.tex, namely 

 --- a class CM, the "trivial" class of Computer Modern for
which E is empty,

 --- a class for TM fonts which (roughly) has E the set of
accented characters in the segment 192-255 of the Cork norm.

 --- a class AD, the Adobe type one font distribution
reencoded to extended CM, August 1992 with N. Jungers.
Coincident with TM in Aug 1992, but hopefully to become
somewhat larger. 

      Other classes can be added with help from a TeX 
programmer.  

      Since for each font, CaesarCM needs to have a specified and
known extension class E, we have to provide means to specify it.

     1)  By default the assigned class will be empty, as for CM
fonts, but this can be changed to TM (say) by the command
\let\CaesarDefaultClass\CaesarClassTM.

     2)  The .tfm file may carry external or internal markings that
cause CaesarCM.tex to conclude automatically that the font belongs
to a certain class.  See the following appendix for details.

     3)  The class can be (re)assigned by command at any time by

      \skewchar<font>=\CaesarClass<class>

where the current possibilities for <class> are CM and TM.  For
example with Plain TeX,

      \skewchar\tenrm=\CaesarClassTM

assigns class TM to \tenrm.  (Currently, \CaesarClassCM gives -10
and \CaesarClassTM gives -20 but these numerical values may be
ephemeral.) Such an explicit assignment takes priority over an
internal marking as in (2) above.

      There may be several font commands, say \tenrm and \myrm that
refer to the same font cmr10; happily the above class change applies to
all at once --- they are indeed the same font.


                   Appendix  MARKING EXTENDED CM FONTS


      CaesarCM uses the notion of a

                  "marked and extended"

CM (metric) font.  The marking serves to automatically specify an
extension class to which the font belongs.  Note that the marking
can always be replaced (or overruled) by an explicit class
assignment as in (3) above.  Marking is thus an optional extra
convenience. 

 *** External marking by name prefixes.

       CaesarCM automatically recognises the following two-letter
".tfm" file name prefixes:

        cm, CM, dm, DM, tm, TM

If one of these prefixes occurs it tests whether <eaccute>
(at Cork position "E9) has positive width.  If so it
assigns class TM. (A standard CM font of course fails this test.)

        This is a test of last resort that applies only if
other tests would say nothing (ie give the default class).

        Norbert Schwartz and Daniel Taupin have recommended accent
management based entirely on prefix tests. (See the June 1992 GUT
forum proceedings including some macros of Schwartz) CaesarCM makes
prefixes one of many useful tests; its use of \skewchar as the basic
prose font class identifier allows this diversity.


 *** Internal marking.

       All fonts distributed with CaesarCM are marked internally
marked to assure that their behavior under CaesarCM is (normally)
independant of the name assigned.

      The internal marking used is very strange; so it is well to
explain first why mundane solutions seem to fail.

      For other TeX packages, the marking device most used is a
distinctive prefix for the font name.  This works because the
filename of a font \myfont is available to the programmer through
the \fontname primitive.  For example, \edef\temp{\fontname\myfont}
places the fontname (without the .tfm suffix) in the expansion of
\temp for further processing. However, to minimize space needed and
also hassle, CaesarCM recommends using extended CM fonts under the
name of a naturally corresponding unextended CM font.  Thus another
marking device is needed.

      Other sorts of software, tend to use headers.  Indeed, most
data structures include headers. The .tfm format (which originated at
Xerox Park) is no exception. Unfortunately TeX seems to heed little
of this information and passes none on to the TeX programmer.

      One can readily store accessible information in extra font
dimensions (see TeXbook page 433). Unfortunately, when one consults a
nonexistant font dimension, an error message is produced.

      Here is the primary `mark' used by Caesar (version 2)

(*) --- (small) positive depth for the ASCII character "5E, the hat
accent.  

      This indicates that the class is not CM.  It also indicates
that fontdimension 8 is present, which then leads to further information
on the class.

      The hat is a character which, under normal typographic variation,
will retain depth exactly 0, since its geometric "bounding box" is far
above the base line. This information cannot be tampered with by other
macro packages. Seemingly, it is reliable inasmuch as every alphabetical
font (including PostScript type) that is likely to be encountered in
prose will have hat in this position.

     You can test whether a .tfm file satisfies condition (*) by
installing it as indicated in the main text, selecting the font in
question, and then using the line

     {\catcode`\^=11\setbox0=\hbox{^^5e}\showthe\dp0}

     While only one font class is in use (the class of TM),
no other information is needed.  However other classes
will be distinguished later on by special values of 
various fontdimensions. For example, it is expected that

(**) --- a small negative value for fontdimension 8 will indicate
that the font class is TM (or AD).

     When other indications are needed the above rules will be
extended.  Contact the author if the need arises.

     Postings of extended computer modern fonts will hopefully be
accompanied by .tfm files that provide the above marking. If
you find that this is not the case, you can use the command 

       \skewcharacter\myfont=\CaesarFontClass<classtag>

Alternatively, here is a way to mark a given .tfm file as being
of class TM. Use a conversion .tfm <--> .pl, as provided in a
metafont installation, to be able to edit the .tfm file in its
readable ASCII .pl form. Then change the .pl version by addition of
a line

       (CHARDP R 0.0001)

to the entry for the character number hex "5E = octal '136. This
makes the depth .001 pt.  (Negative depths are rejected by Knuth's
PLtoTF program.)

     As for (**) you add a line of the form

       (PARAMETER D 8 R -0.0001)

after the line 

       (EXTRASPACE R ???????)

specifying fontdimension 7. 

     Since you can expect to need just a few fonts to obtain adequate
hyphenation this "hacker's approach" should be adequate. 



      APPENDIX. The Smart Accents Mechanism

    There is a fragility of the TeX kerning and ligature mechanism
that is a serious limitation on the way this CaesarCM's announced
behavior can be programmed. It is a surprising fact that TeX code
intervening between two characters strongly tends to prevent kerning
(and ligature) of the characters.  Indeed any code that does not
evaporate without trace under macro expansion (in TeX's mouth) seems
to prevent kerning.   For example, \count255=1 and {} both leave some
trace, whereas \if\count255=0\count255=1\fi leaves a trace only if
count register 255 is 0, and \empty never leaves any trace.  The
rather innocent looking traces mentioned here do not interfere with
hyphenation. But they do prevent kerning (and ligature), as Bernd
Raichle <raichle at azu.informatik.uni-stuttgart.de> pointed out on
the GUT list (July~1, 1992). In version~1 of CaesarCM, kerning
between a ready-made accented character and the character preceeding
it was hopelessly poisoned by spurious typesetting designed to test
whether a ready-made accented character was present in the current
font.

     Unused potential of the \skewchar mechanism of TeX provided means
to skirt the above difficulty.   First some background. There is an
integer font parameter \skewchar<the font> which is automatically set
to \defaultskewchar, normally -1, when a font is loaded. For some
math fonts (e.g., cmsy but not cmmi), Knuth sets it in Plain TeX to
an appropriate (positive) character code that then serves to control
accent placement in math mode. It seems that \skewchar has never been
exploited for prose fonts.

     The idea is to assign for each prose font a large negative value
(one less than \defaultskewchar) to the function \skewchar that will
serve to identify the extendion class of the font.

     Here is, in outline, how an accent  macro like \'#1 then works.
A `current extension class' integer register is first compared with
\skewchar\font.  If they coincide, the current accent handling macro
is used. When extension class TM is current, this macro checks whether a
ready-made accented character has been assigned to the triple
TM,',#1; this is an \ifx...\fi test that leaves no trace if the
answer is yes. If yes use it, and if no use the recipe of HyAcc-CM to
form a composite character.  This much runs at high speed, and in the
case of ready-made accented characters, it allows kerns and
ligatures. In the rare case when \skewchar\font is distinct from the
current extension class, CaesarCM goes into a more time-consuming setup
subroutine as follows. If \skewchar\font is still the initial value
\defaultskewchar, then, as a preliminary step, determine the extension
class as indicated in the two appendices on extended CM fonts, and
reset \skewchar\font to this value, ending the preliminary step. Now
reset the current extension class to \skewchar\font and \let the current
accent handling macro be one predefined for that extension class.

     

  --------------------------



APPENDIX Oz.  MACINTOSH COMFORTS FOR OzteX USERS.

     For Macintosh users, some documentation is provided in a reader
application.  This application should not be ramoved from your TeX volume
because it carries icons useful for identifying font files for
CaesarCM.

     Namely the bitmap and .tfm files for the CaesarCM mechanism get
highly visible icons (distinct ones for OzTeX and for Textures).  Other
.pk's and .tfm's are unaffected. This originally applies to the the
.pk's and .tfm's provided in the CaesarCM package for OzTeX. To
maintain this behavior for other files you later introduce, you should use
the universal Macintosh system tool ResEdit to change the type and
creator strings respectively to

        **pk   and  '`^"

        *tfm   and  '`^"

For this see "info..." in the file menu of ResEdit. Under some
Macintosh systems it may be necessary hold these files on a diskette
for the change.  In all cases, to bring the new icons into view, the
documentation reader should be kept present alongside the files.

     These icons make it clear for which fonts you already enjoy
enhanced hyphenation.  They do not, at present, reveal the degree of
enhancement, but they may in future versions.

  --------------------------


APPENDIX Textures.

      Postings of the CaesarCM package include some extended CM
fonts for Textures, the first implementation of TeX on the
Macintosh.

      Macintosh format font data should reside in files of type
FFIL; they carry distinctive suitcase shaped icons. The data is
automatically accessed by Textures if it is placed in a folder
called TeX Fonts  which lies in the same folder as Textures.
Suitcases can also be accessed from within Textures and through
Suitcase II.

      The .tfm files for Textures are in Macintosh resources of
type "*TFM". The contents of the resource coincides exactly with the
ordinary .tfm file.

      Bitmapped fonts for Textures are of Macintosh norms rather
than .pk. Each is of one of the following types.

      (a) resources of type "FONT", possibly accompanied
          by a resource of type "FOND"

      (b) resources of type "NFNT", possibly accompanied
          by a resource of type "FOND"

Macintosh and Textures font administration seems to require that
all magsteps for a given font be present in the same file and be
(correctly) described by a single "FOND" resource.

       Think of a complete font like cmr10 as a triple: a *TFM, a
FOND, and thirdly a bunch of bitmaps either of type FONT or NFNT
--- essentially one bitmap for each magstep available, but a
few extra for the preview.


MAINTENANCE AND TRANSPORT of fonts for textures for Textures

     In practice, an absolute necessity for the administration of
Macintosh bitmapped fonts is the utility Suitcase II and its
sidekicks Font Harmony and Font&Sound Valet.  The last of these
provides compression --- twofold or better, which essentially
corresponds to the passage in the TeX world from .gf (uncompressed)
to .pk (compressed) . Suitcase II provides access to the compressed
files. (The .pk compression is superior.) Font Harmony and Font&Sound
Valet provide administration; in particular the maintenance of
suitable FOND resources. Suitcase II costs about $50.  So long as
there is no competitive alternative to the use of Suitcase II,
expect to find the extended fonts for Textures posted in Suitcase
compressed format; such compressed fonts will not work at all
without Suitcase.

     The inferiority of Suitcase II's compression relative to
Rokicki's ".pk" compression means in practice that the cost of TM
extending cmr10 is 75K rather than 40K --- for all seven magsteps;
for magsteps <=2 it is 32K rather than 17K.  Users of Textures with
disk space problems often restrict to magsteps <=2.

     Use ResEdit for inspecting and editing all these font
resources. Use it also to move all except the FOND-NFNT combination
of of type (b) above. The latter job (only) should be left to
FontDAMover.

     Exercise great caution in employing FontDAMover, particularly if
any FONT resources are present:

   (a) Use only a very recent version of FontDAMover, and if you
work under under Multifinder, assign it a maximum partition.

   (b) First use Font Harmony to eliminate incoherencies.

   (c) If FONT resources are present, first use ResEdit to remove
them and their associated FOND's to another suitcase file.

   (d) Keep suitcase files down to moderate size and keep backup
copies.  Move just a few fonts at a time.

Note that any FOND-NFNT combination that is moved with
FontDAMover will be decompressed by the move.  As compression is
often extremely slow, it is better to replace such moves by
copying and FontDAMover deletions.


INSTALLATION AND REMOVAL of extended Mac fonts.

     Although the mechanics differ, the principles are exactly the
same as for a generic TeX installation.

     Installation of extended fonts whose names do NOT coincide
with fonts already installed is a trivial matter. Just push the
suitcase containing the new fonts into the "TeX Fonts" folder
residing in the same folder as Textures. For fonts supplied with
CaesarCM this suitcase includes the relevant *TFM resources.
Removal is the obvious inverse provess.

     In case they DO bear the names of CM fonts they extend, your
extended fonts in effect replace the pre-existing fonts of this
name. But, for purely administrative reasons, it is better to keep
the extended fonts in their own suitcase.  Then installation of the
extended fonts neither removes nor destroys the font of the same
name it is replacing.  That is the users responsibility, and
failure to attend to this will cause terrible chaos!

      Later removal of extended fonts is really the same process as
installation (you replace the extended by the nonextended).

     Remember to attend to BOTH the .tfm's and the bitmapped fonts.
Use font Harmony after every installation. Remember to recompile
your format when the fonts it uses are modified.

     If you encounter strange behavior in spite of the above
precautions, consider gradually rebuilding your font collection
starting with the extended CM fonts under Plain TeX. Many
collections are somewhat raggedy and need a rebuild after years of
use.  Half of those I have seen are in this category.

     CaesarCM is fully compatible with the CM/PS Postscript norm CM
font collection marketed by Blue Sky Research; indeed bitmapped
fonts in principle can be freely mixed with them. However this sort
of configuration is the most complex in service and has the most
bugs.

      Be alert for the following standard ailments that only
ResEdit can detect: missing *TFM resources, duplicate *TFM and
NFNT resources. 

CREATION OF BITMAPPED FONTS for Textures

    Blue Sky Research has permitted the posting on

   ftp rusinfo.rus.uni-stuttgart.de  [129.69.1.12]

of a collection of Macintosh Programmer Workshop (MPW) tools for the
conversion of ".pk" and ".tfm" resources to equivalent FOND, NFNT, and
*TFM resources.   If you start from "marked extended CM fonts" for a
generic TeX, you get after conversion the same for Textures.

   There are two technicalities to note.

(a)  There are .tfm files in some versions of Textures notably 1.4.
These appear to be relevant to the use of fonts of PostScript norm
only; try to ignore their presence; removal or tampering will cause
PostScript fonts to misbehave.

(b)  Currently, *TFM files should be adjusted using the EdMetrics
application to have
   --- Quickdraw name equal to their resource name,
   --- style Plain
   --- Quickdraw map of "Other" type, number 6
 Check current distribution for possible changes!



Appendix.  DESARMENIEN'S PIONEERING WORK.

      Although this package was inspired by the availablity of the
extended CM fonts of D. Taupin, it might reasonably have been
inspired by rereading Jacques D'esarm'enien: How to run TeX in a
French environment: hyphenation, fonts, typography, TUGboat, 5
(1984) 91-102.  The key concept of a special font class with
ready-made accented characters occurs, and this includes a
membership test, albeit one acceptable *only* for French, namely
small negative value for fontdim 7 (the extra space between
sentences). With Desarmenien's macros, \`#1 is redefined  as
follows: if the current fontdimension 7 is negative and the macro
argument #1 is e, then use a tailor-made e-grave, and otherwise
use Knuth's original accent macro. Desarmenien remarks that he
would have used extended CM fonts but for a limitation of the 1982
version of TeX, namely the impossibility of having true letters
(with \lccode positive) in positions >127. He gives credit for
an unspecified subset of his TeXnical tricks to Don Knuth and John
Hobby.

   


Laurent Siebenmann, Mathematique, bat 425, 91405-Orsay, France
        siebenmann@rsovax.ups.circe.fr (tex)
        siebenmann@LALCLS.IN2P3.FR

POSTINGS: ftp rsovax.ups.circe.fr  [130.84.128.100]
           ftp matups.matups.fr  [192.54.146.4]

 Version 2 of Sept 1992 (alpha)