<!--tim_hist.txt: History of changes since draft version 2.4 of the TIM DTD.

     Old versions of the DTDs are archived off-line. If you should ever have a
     need for one, request it from dfp@ims.bellcore.com.

     Changes that almost certainly require updating of older document 
     instances are marked with "!". Some changes have been completely ("*")
     or partly ("+") undone in later versions.

     Last update: 1995-04-30, for Release 1.1.10.

     In Release 1.1.10, an entity &smark;, for a service mark (SM), was added
     to tcif_kb.ent, timm.dtd, and timm.hdr. (There are now 2 entities in
     tcif_kb.ent that aren't for keycaps - the other is &newline; - but there
     is still, deliberately, no entity for tab, which, if ever encountered in
     a TIM file, should be treated as collapsible whitespace.) The change does
     not affect existing TIM instances.

     In Release 1.1.9, the content models of OrderedList, UnorderedList, 
     VariableList, and SegmentedList, and their ...ListGrp subelements were
     modified to allow Annotes and Danger, Caution, and Warning messages
     immediately after any title for the list or group. This allows a note
     or admonishemnt that applies to a whole list or list group to occur
     where it logically belongs in the hierarchical structure. The changes
     do not invalidate any existing TIM instance.

     In Release 1.1.8, notations were added for DXF and PDF files. The
     changes do not affect existing TIM instances.

     In Release 1.1.7, Section was added to the content models of Contents
     and Index so that headings can be used within provided lists. And a
     notation was added for PNG (Portable Network Graphics) files. The
     changes do not affect existing TIM instances.

     In Release 1.1.6, notations were added for Animator (.flc & .fli), AU
     audio, AVI video, and WPG graphic files, anticipating new testing before
     establishing TCIF Guidelines on graphics and multimedia formats. The
     changes do not affect existing TIM instances.

     In Release 1.1.5, column headings (MarkTitle and/or TermTitle, and LITitle
     or SegHead) previously available in ...ListGrp elements were added as
     options for the parent block-list elements (UnorderedList, OrderedList,
     SegmentedList, and VariableList), so that the headings could be specified
     for an entire list. (Not enabling this was an oversight.) Also, two
     notations were added, for black-and-white TIFF with CCITT Group 4
     compression and for WK1-format spreadsheets. Group 4 is a useful
     compression scheme, but it is not understood by most TIFF file viewers; a
     separate notation allows a different viewer to be specified for it. WK1 is 
     a lowest-common-denominator spreadsheet file format. Finally, the close
     attribute of Pt has been changed from NAMES to PCDATA, to match the remap
     attribute, as it was supposed to. None of these changes will invalidate
     any existing TIM instance.

     In Release 1.1.4 - well, it turns out that GlossEntry did need a change
     (see entry for 1.1.3), from LI+ to LI*, so that a "See" entry would
     make sense and format cleanly. No TIM instances will be invalidated.

     In Release 1.1.3, the content model of IndexEntry was expanded to allow
     multiple Secondary and Tertiary elements, needed if IndexEntry is used
     in marking up the content of an actual index. This does not invalidate
     previous TIM instances. No such change is needed to allow GlossEntry
     to be used in an actual glossary. The original intent was that these
     elements would be used to store the raw material for generating
     glossaries and indexes; this change may reduce the processing burden on
     both originators and recipients when an index already exists.

     Release 1.1.2 changed the content models of SegListGrp and DistSegLI to
     make them recognizable as tables to typical current applications - that
     is, a wrapper was added around the LITitles (and, for DIstSegLI, also 
     around the LIs) so that elements can be identified as Table (SegListGrp
     and DistSegLI), Row (SegLI and the new SegHead), and Cell (LI and
     LITitle).  While this change invalidates a small number of existing
     document instances, it is being treated as a bug fix, since it makes two
     element types that were destined to be neglected into something
     potentially useful. If you've used SegmentedLists with LITitles as column
     headings, surround each set of LITitles with <SegHead>...</SegHead>.
     If you've used DistSegLI, do the same and also wrap the LIs in <SegLI>...
     </SegLI>. A second "bug fix" is removal of the altreps attribute from Pt
     and restoring it to Span. The effect is to require that, when there is an
     alternative representation of a virtual structure, both the start and end
     points of the preferred representation be identified.
     Also in this release, the attributes in attribute declarations were
     rearranged into the preferred order of occurrence in element start-tags
     - it is recommended, not required, that attributes occur in the order
     you see them listed in timm.hdr.

     Release 1.1.1 made three minor changes to 1.1.0. SimpleList was excluded
     from the content model of ListTerm (this prevents nesting of SimpleLists,
     as had been intended), and the default values of two HyTime attributes
     were changed to reflect the likeliest processing defaults: NmList's
     nametype attribute was changed from "entity" to "element", and DataLoc's
     quantum was changed from "word" to "norm". There were also minor changes
     to the wording of the copyright notice.

     Release 1.1.0 is the version actually submitted to TCIF letter ballot. It
     differs from Release 1.0.7 in the following ways:
   ! The ISO-GRK3 entity set (appropriate for using Greek letters as symbols)
       was substituted for the ISO-GRK1 set (intended for Greek prose).
     The spanend attribute was removed from all elements except Span and
       IndexTerm (effectively requiring that Span be used to mark the starting
       point of any virtual structure).
     The name of the preferred attribute on IndexTerms was changed to prefer,
       so that the value "preferred" can be used alone.
     Support for HyTime linking was improved: Tree locators and data locators
       were added to the existing name locators. All HyTime links should use 
       the Link, BiLink, or Ref elements. URLRef should be used for links to
       World Wide Web URLs, and DBRef and DBLink should be used for
       non-HyTime, non-URL interdocument links. BiDBLink was eliminated.
     Many small errors and inconsistencies were corrected.
   ! The version number was reintroduced into the public identifier: 
         "-//USA/TCIF//DTD TIM-1//EN" (for the standard version, tim.dtd)
         "-//USA/TCIF//DTD TIMM-1//EN" (for the "minimal" version, timm.dtd,
           which you should use if you need no extensions, such as non-standard
           elements, attributes, or character sets)
       As with the draft versions, the main version number will change only
       when changes to TIM are significant enough to make most previous
       document instances invalid (that is, a typical TIM-1 document instance
       wouldn't parse using a TIM-2 DTD). However, such changes to TIM should
       now occur in time intervals measured in years rather than months, and
       will always be accompanied by explicit scripts (e.g., a perl or awk
       program) to update old document instances.
       The second part of the version number will be incremented when there are
       enhancements to TIM that allow new elements or attributes to be used.
       The public identifier will not change because the new TIM (e.g., 1.2.0)
       will still work with older document instances (e.g., those made for
       1.1.0). However, style sheets and conversion programs will need to be 
       updated to handle the new features. The third part of the version number
       will change for bug fixes or other minor changes that may not require 
       any changes to processing.
       The general rule: old TIM-1 document instances will parse if you use the
       latest TIM-1 DTD. (It is not always true that old TIM-1 DTDs will allow
       the latest TIM-1 instances to parse - you should always use the latest
       version of the DTD.)

     Release 1.0.7 corrects the way entities are used to allow content-model
     extensions: an entity used directly in a content model must contain a 
     token, according to the strict interpretation of the SGML standard 
     followed by the sgmls parser; so some content-model definitions were made
     indirect to let them include entities with default definitions of "".
     Meanwhile, some further reduction was made in superfluous attributes. 
     For instance, contents was removed from all elements except AltRep,
     spanend from structure elements (spanend being used to define a virtual
     structure where an actual one is illegal), and emph from TelDoc.  End-tag
     omission was prohibited for non-empty inclusion elements (Float, AltRep,
     etc.), to avoid problems if anyone uses TIM with OMITTAG set to YES in
     the SGML declaration.

     Release 1.0.6 corrects the inadvertent omission of an id attribute from
     the subelements of IndexEntry. The ID is needed when the contents
     attribute is used on another such subelement.

     Release 1.0.5 differs from release 1.0.1 (95/10/30) in the following ways:
     Table was added to the content model of AltRep (omitting it was a mistake)
     and then excluded from Title and literal-text elements.
     Another error was corrected in the exclusion group for titles, which
       included Annotes (and therefore excluded them from titles).
     Some minor inconsistencies in the three versions of the DTD were removed.
     The third version is new, and the others have been renamed since draft 10.
       They are:
       * tim.dtd, the reference version, with public identifier
           "-//USA/TCIF//DTD TIM//EN" 
       * timm.hdr, which incorporates tcif.dcl, calstabl.dtd, tcif_kb.ent, and
         the 7 ISO entity sets referenced by tim.dtd to simplify processing
         with some applications, and expands all parameter entity references
         within tim.dtd to their default values. This makes the DTD more
         readable but also removes the possibility of extending or modifying
         the DTD by normal means (since no extensions have yet been agreed upon,
         this may be seen as an advantage). The extra M is for "minimal." No
         public identifier is assigned to this file.
       * timm.dtd, which incorporates the same files as timm.hdr except for 
         tcif.dcl, and also has parameter entities expanded to their defaults,
         but has all comments and extra white space removed so it can be
         retrieved over networks and loaded into applications in about 1/3 the
         time of the reference version. Its public identifier is
           "-//USA/TCIF//DTD TIMM//EN"
       Document producers are encouraged to reference TIMM (timm.dtd) for all
         documents that do not require extensions to TIM, so that recipients
         will know that they can use default processing.
     All 3 versions incorporate the new, official ATIS copyright notice, and
       all 3, almost by coincidence, are even multiples of 100 lines and 100
       characters in length, which has the minor advantage of making most
       modifications easy to detect.

     Release 1.0.1 differs from release 1.0.0 (95/10/23) in the following ways:
     The attribute alerts was added to most elements to use as as an unequivocal
       link to Alerts (presumably containing admonishments). This and the next
       two changes have the goal of making mandatory display of admonishments
       easier to implement.
   + The Spanend attibute was added to most elements (including Point, which
       can be inserted anywhere) to reference the end of a span that begins
       with the element using spanend. This makes spans more concrete, and,
       when content-bearing elements begin and end the span, it lets
       unsophisticated browsers provide a larger hotspot for linking to an
       admonishment.
     The Span element was redefined and redesigned. It is no longer to be used
       to indicate any span, since spanend is available for most elements. Its
       use is now limited to a virtual structure whose attributes apply to all
       content between it and the span-ending element. In other words, it is
       to be used only where the hierarchical structure does not allow a single
       element to encompass the range of content an attribute applies to, as
       when change bars are applied from the middle of one paragraph to the
       middle of the next. The new design of span is a point element with a
       single required spanend IDREF, rather than a floating element with two
       linkends.
     The DBIndexEntry element was found to be unnecessary. IndexEntry now has
       both the rid and dbid attributes for intra- and inter-document Xrefs.
     Minor errors in assignment of attributes to elements were corrected, the 
       entity declarations for content models and attributes were reorganized
       to make them easier to use, and more comments were added.

     Release 1.0.0, intended to be the version of the TIM DTD submitted to TCIF
     letter ballot, differs from draft version 10.0.0 (95/08/03, the beta
     version) in the following ways:
   ! The elements FrontMatter, Body, AppMatter, and BackMatter were added as 
       divisions of the content of TelDoc after the DocID. Each may occur
       once, and Body is required. All content of TelDoc after the DocID may
       be enclosed in Body, or it may be distributed into 2, 3, or all 4
       parts. These elements help the current generation of applications with
       their autonumbering (most style sheets will probably assume that child
       Sections within Body should be numbered and those with AppMatter should
       be lettered).
   ! For similar reasons, the content models of LI and DistOrdLI were changed
       to eliminate PCDATA, making a P or other subelement required rather than
       optional. This will give browsers consistent input so typical formatting
       rules will work with all doc instances.
     Another change for the sake of current tools, but only to instructions:
       The format of Listing should be assumed to be "verbatim", and that 
       value should not be specified explicitly (because doing so will cause
       some current tools to treat the contents as if "CDverbatim" had been
       specified). "Verbatim" content should be parsed - that is, entities and
       subelements (like Emph) should be recognized, but white space and line
       breaks should be preserved.
     Float was split into Float and Alert elements, and the alert attribute for
       Float eliminated. Again this was done with the thought of simplifying
       implementation in existing applications.
     The AlternateText element, which would not work as previously defined,
       was redefined and renamed AltRep. AltRep can contain almost any elements,
       and those with PCDATA in their content models can be in a format
       (NOTATION) other than TIM.
     SimpleList was given numbering attributes, for simple lists 
       with items that are (1) numbered, (2) lettered, or (3) otherwise 
       enumerated, as in this sentence. SimpleList was also given the
       attributes separator and lastsep so that the separators in an in-line
       presentation of the list could be specified.
   + The element HotSpot, added to draft 11 (never released) to provide a way
       to specify hotspots within Frames and Graphics enclosed in Frames, was
       left out of 1.0.0. The explanation for HotSpot, preserved here in case
       it is restored: Placement can be specified with
       new attributes provided for Frame, Graphic, and HotSpot. For Frame and 
       Graphic, width and height can be specified in pixels or in units of
       measure using the dimensions attribute. A Graphic, HotSpot, or sub-
       Frame can be positioned with the placement attribute, whose 
       measurements should place the center of the current element to the
       right of the left edge of the enclosing element by the amount specified
       in the first value and down from the top by the amount specified in the
       second value. (That explanation doesn't read well but it describes the
       most straightforward way to position a hotspot graphically.) 
     The format and sortas attributes were removed from elements that could
       not contain PCDATA. The id and label attributes were removed from 
       titles and a few other elements. Here and there, other attributes 
       that were impractical with certain elements were removed from them.
       Consequently the number of "base attributes" has been reduced.
     The attribute scope was added to the Contents and Index elements, to
       allow specification of the enclosing element types or the range of a 
       document from which to collect entries.
     The altreps attribute, inadvertently omitted from Graphic, Object, and
       ExternalText, was added to them.
     The position attribute for numbered elements, an ill-considered foray
       into specification of layout, was dropped.
     "Holds" was added as a value defined for the continuation attribute, so 
       that split numbering could be specified more explicitly.
     "RTF" and "HTML" were added to the list of NOTATIONs available for 
       elements (notations that can occur within a TIM file).
       "EPS" was removed.
   ! The version of TIM in the required Status element should now be 1.0.0.
       (The earliest draft made public was 2.4, and as far as is known no
       TIM document instances were created before Draft 8, so no confusion
       is expected from the change in version numbering.)

     Draft 10.0.0 is the final "beta test" version of the TIM DTD. It
     differs from 9.0.1 (95/06/28) in the following ways:
   + Index and glossary elements were simplified. IndexEntry now has an
       IDREFS attribute to reference any element or span of elements. 
       IndexTerm now has only an ID, not IDREFs to IndexEntrys, and its 
       use is optional. Because there's no IDREF in IndexTerm, DBIndexTerm 
       was eliminated, but DBIndexEntry became necessary. GlossTerm and 
       DBGlossTerm were eliminated (use Ref and DBRef). GlossEntry remains.
     The attribute target was added to Ref and DBRef to specify the target 
       element's type, as an aid in generating appropriate wording. 
   ! Numbering streams (especially for Frames and distributed-list items) 
       will now be identified through additional values of the type 
       attribute rather than through the group attribute, so that only one 
       attribute value needs to be checked. This is a change not to the 
       DTD but to the the "data dictionary" in the documentation.
     Some parameter entities have been added to make more attributes 
       redefinable, and some entities have been renamed for consistency.
     Elements and attribute values likely to be used only in telecom 
       documents were defined as local modifications. The element Admon was 
       restored for general use.
     The %TCIFtype; entity set was eliminated. A convention for indicating
       typographic changes in verbatim text will be documented.
   + The DocID element in a TIM file must now have at least one Status 
       element. It should identify the TIM version used to encode the file: 
       <Status>TIM-10.0.0</Status>. (The version should be at least 10.0.0.)
       This element is allowed but not required in a TEDD package's DocID, 
       as defined in IPI-95-001.

     Draft 9.0.1 is the "beta test" version of the TIM DTD. It
     differs from 8.0.7 (95/06/16) in the following ways:
   ! The ID attribute was removed from Float. The ID should be on each
       substructure the Float contains.
   + IndexTerm and GlossTerm were redesigned. The contents of the index/
       glossary entry (what goes in the back of the document) are now in
       IndexEntry and GlossEntry, which may occur anywhere in the file.
       The terms in the text that are linked to those entries should now
       be marked as IndexTerm/GlossTerm if the ..Entry is in the same file,
       DBIndexTerm/DBGlossTerm if it is elsewhere.
     Several notations were changed from SYSTEM to PUBLIC as public 
       identifiers were found for them.
   * The %TCIFtype; entity set was added to provide a way to indicate
       typographic changes in verbatim text.

     Draft 8.0.7 is a revised "alpha test" version of the TIM DTD. It
     differs from 7.1.0 (95/03/13) in the following ways:
   ! The DocID element and its subelements were copied from the DocID DTD, 
       and DocID was made the required first element within TelDoc, 
       replacing TitleGroup (it includes a TitleGroup). In a TEDD package,
       identifying information will now appear within the TIM file as well
       as in the docid file. 
   + The Admon element was replaced by Danger, Caution, and Warning, the 
       standard types of admonishments in telecom documents.
     The GlossTerm element, similar to IndexTerm, was added. 
   ! The AltRep element was replaced, and Graphic redefined. There are now 
       AlternateText and GraphicalText, both with content, and ExternalText,
       Graphic, and Object, all empty elements with required entity refs. 
   *   (The altreps attribute may refer to any element.)
     The empty attribute was removed: element instances that happen to be
       empty (other than those empty by definition or by virtue of an entity
       reference using the entityref attribute) must now have endtags.
   ! The emph attribute was given the values formerly assigned to the type 
       attribute of the Emph element, and for Emph it replaced type.
     The inheritnum attribute was given new values, and the inheritfrom
       attribute added, to allow precision in specifying inheritance of
       element numbers.
     Listing was given the attribute width to allow specifying of the
       optimal number of characters per line in displaying literal text.
       The attribute widths was renamed segwidths to avoid confusion.
   + The attribute spanend was made IMPLIED rather than CONREF. It should
       be used in a non-empty IndexTerm and reference an element at the
       end of the page range for the index entry, rather than being in
       an empty IndexTerm at the end of the range referencing an IndexTerm 
       at the beginning.
     The attribute rid (an IDREF) was added to Primary, Secondary, Tertiary,
       See, and SeeAlso, so that the content of the element could be copied
       from another such element that it should match (this may prevent
       spurious index entries due to typos).
     The sortas attribute was made a base attribute, so that any element
       that might be collected into an index could be given special sorting.
   ! The ssmap attribute was given the user-friendlier name type, and dsmap
       was renamed dstype. The maps attribute of Contents and Index was
       consequently split into types and remaps. Several parameter entities
       were also given friendlier names.
     The include attribute defined for Contents/Index was added to Ref and 
       DBRef to allow specification of a reference's text content.
     References to the entity-declaration sets for special characters were
       moved from the invocation file into this file for convenience.
     The attribute close was added to Pt so that explicit info could be
       passed on alternative-structure elements closed at that point.

     Draft 7.1.0 is the "alpha test" version of the TIM DTD. It differs
     from 7.0.2 (95/02/27) in the following ways:
     The ssmap, group, and emph attributes are defined as NAMES rather than
       CDATA, to make the values case-insensitive. 
     The Rep and Link elements have been further redefined to comply with 
       the HyTime standard. The changes are only to defaulted attributes.
     The elements NameLoc and NmList have been added for the same purpose.

     Draft 7.0.2 differs from 6.0.2 (2/20/95) in the following ways:
     The facets attribute was removed. It may be restored when a
       faceted-classification scheme has been developed.
     Separate entities were defined for the ssmaps of most elements, so
       they can be redefined easily in TIM extensions.
     The DocBook elements for index terms were added (with slightly
       simplified content models and attribute lists). A sortorder
       attribute was added to Index.
     Rep, Link, and Span were redefined to comply with the HyTime standard.
       Rep is now Rep and DBRep (both clink), and Link is now Link, DBLink,
       BiLink, and BiDBLink (all ilink, the Bi... versions bidirectional).
       Span is now an ilink.
     Several NOTATIONs were added for additional data types.

     Draft 6.0.2 differs from 3.0.5 (1/3/95) in the following ways:
     The copyright notice was changed to make TCIF coholder.
     End-tag omission ("- O") was specified for all elements because some
       parsers require that elements with CONREF attributes be declared
       that way. (The TCIF SGML declaration doesn't allow end-tag omission.)
   + The attributes keywords and facets (faceted classification) were added
       to the base attributes. Keywords may be any terms useful in searching.
       Facet terms must come from a restricted vocabulary (to be developed)
       that provides the equivalent of a multi-faceted library classification
       of subject matter; i.e., each term is the verbal equivalent of a Dewey
       decimal number.
   ! The element Title.comp was renamed TitleGroup, and content models that 
       included (Title|Title.comp) were changed to (TitleGroup). This means
       that for all elements except list items, Titles are always (instead of
       optionally) enclosed in a TitleGroup (a TitleGroup always contains
       exactly one Title). The slight increase in verbosity is more than 
       offset by the simplification in processing for both encoding and 
       rendering applications - they have one alternative to allow for
       instead of two.
     The elements Contents and Index were added.  They have the same content 
        model as Sections but cannot contain Sections (there could be multiple 
       Contents or Indexes in a Section). Each may contain already-generated
       content (presumably Refs containing links rather than typed page
        numbers) or may be an empty element whose attributes give instructions 
       for generating the content. For example, <Contents elements="Section" 
       levels="1 2 3 4 5" include="number title link"> means generate a table
       of contents listing the section numbers and titles of the first five
       hierarchical levels of Sections, and generate a link from each entry
       to the Section itself. <Contents elements="Frame" groups="Table" 
       include="number title link"> would request a comparable list of 
       tables. <Index elements="T" maps="ndx" include="text link"> requests 
       that the textual content of T phrases identified by ssmap="ndx" be
       listed alphanumerically. Contents and Index are both allowed anywhere
       in the content of TelDoc or Section (or Rep), and have identical 
       content models and attributes, so the only difference is that items 
       of an Index are ordered alphanumerically and items in Contents are in 
       the order they appear in the document.
   ! The content models of OrderedList, UnorderedList, SegmentedList, and 
       VariableList were changed to require OrdListGrps, UnordListGrps, etc.
       (renamed from OrdListSecs, etc., for consistency), as containers for
       the individual LI's. This has three advantages: simpler processing, 
       as with TitleGroups (above); consistency with CALS Tables, with which
       many lists would be interchangeable; and the freeing up of the list's
       numbering attributes to apply to the list itself, so it can be 
       numbered, as all comparable structures (Admons, Frames, etc.) can be.
     The content models of DistSegLI and DistVarLI were corrected to be
       analogous to (not the same as) SegLI and VarLI. The main difference is
       that titles can be specified for the columns, as in SegListGrp and 
       VarListGrp.
   ! The elements LI and LI.comp were combined for simplicity, with a mixed-
       content model (mdc) defined to describe the content of the new LI.
     The present attribute for SimpleList was renamed arrange and its default
       changed from Vert to #IMPLIED, to allow the attribute to be omitted 
       from tags even if SHORTTAG is set to NO. 
   + The %Resources; inclusion exception in the definition of TelDoc was
       split into %Float; (Float element, see below) and %Altreprs; (AltRep)
       and the inclusions were grouped in an %idc; entity for redefinability.
  !+ The Alert element was renamed Admon, and alert was made an attribute
       of Float (see below), to disambiguate semantics and presentation.
   ! The FNote element was split into Annote (any kind of note) and Float,
       a container for any structure (S/Admon/Annote/Frame) that floats. The
       float attribute was removed. Float has an alert attribute to specify
       whether presentation of the contents is optional or mandatory.
     The id in Span was made optional so that Span could be used merely to
       mark an alternative hierarchical structure.
     Repset, Rep, and the repset attribute were replaced by AltRep and the
       altreps attribute. Any element may specify alternate representations
       in addition to or instead of content. AltRep is an empty element with
       an entity reference to external content.
   ! The attributes markprefix and marksuffix were renamed prefix and suffix,
       because they are used with numeration but not the mark attribute (they
       were removed from unordered lists).
     The attributes infix and splitnum were added for numbered elements. 
       Infix specifies the string (e.g., "." or "-") used to separate 
       concatenated numbers when inherit=inherit. It goes with prefix 
       and suffix. Splitnum (splitarabic|splitlalpha|splitualpha|
       splitlroman|splituroman|splitoutline) means numbering at a given 
       nesting level should be split into subnumbers: For instance, if the
       third and fourth items in a 5-item list with numeration=arabic have
       splitnum=splitlalpha, numbering will be 1 2 3a 3b 4. (If the items
       were in a sublist, the numbering would be 1 2 2a 2b 3.)
     Listing was given numbering attributes (but not a title).
     Some content models were redefined. In particular, S was removed from
       the %tdc; (text-phrase) content model.
     Element decalarations were regrouped more logically.

     Draft 3.0.5 differs from 2.5 (12/2/94) in the following ways:

     TelDoc:
   ! TelPractice was renamed TelDoc to more accurately reflect the
       applicability of this DTD.

     Sections:
   ! Section1-Section5 were replaced by a nestable Section element to
       increase generality and reusability (this allows sections from
       one document to be incorporated into another without having to
       retag section levels; it also means that end tags cannot be omitted).
     As a consequence, P.comp (a titled paragraph) was removed. A
       titled paragraph should be tagged as a Section or S.
     As a further consequence, the definition of Section content reverted
       to that of Draft 2.4: ((Title|Title.comp)?,(Section|%sdc;)*).
     The content model of Rep was expanded to include Sections, so a Repset
       may now contain alternative representations of the entire document or
       any element of it.

     Lists and Numbering:
   ! Section/Para/Frame/List Marks/Numbers were redefined. The List
       element was split into OrderedList and UnorderedList, and the
       Mark attribute was split into Mark (CDATA), which is only for 
       UnorderedList, and Numeration (None|Arabic|Upperalpha|Loweralpha|
       Upperroman|Lowerroman|Outline), which is for other lists except 
       SimpleList (None is not an option for distributed lists, and Outline 
       is an option only for Sections and distributed and ordered lists). 
       Label remains as a way to specify a particular value for a mark.
     ListSec was replaced by OrdListSec, UnordListSec, VarListSec and 
       SegListSec so that an appropriate content model could be specified
       for the subsections of each type of list.
     DistVarLI and DistSegLI were added as alternatives to DistOrdLI (which
       was DistLI). Analogs for a distributed unordered list and for list 
       sections were not created at this time: unnumbered distributed items 
       would probably never be seen as a list, and list sections would 
       require that the distributed items within each section remain 
       together, which is contrary to the intended use of distributed lists.
       For a distributed list, the information provided by list-section 
       titles could be provided in introductory text, and the option of 
       redefining list marks exists within the items themselves, although
       such redefinition probably should not be allowed, since consistent 
       numbering may be the reader's only clue to the identity of the list.
     The DistList element and the required IDREF attributes for 
       distributed-list items were eliminated, to avoid problems when 
       sections containing distributed items are copied between documents.
       The group attribute can be used to identify the list to which an 
       item belongs. Although consistency in the group attribute cannot be 
       enforced by the parser as IDREFs can, the new scheme does not 
       impose a new burden on authors or processing systems, which can now
       handle - e.g., algorithmically assign numbers to - distributed-list 
       items the same way they handle frames for figures, tables, etc.
       (There can still be a parser error if DistSegLI items copied into
       a document don't include the one that specifies the number of
       items, i.e. segments/columns, in the list.)
     The items attribute of segmented lists was renamed segments to avoid 
       ambiguity in the documentation.
     The widths attribute of segmented lists was redefined so that values 
       can be expressed like values of colwidth in the CALS Table DTD.
 
     New text-phrase elements:
  !+ The text-phrase elements Emph and Keycaps were added. Entities have 
       been defined for common key caps that require more than one keystroke
       to name, such as SHIFT and F1. Text tagged as <Keycaps> should be
       rendered to look like a keyboard key. Emph has the attribute
       type, which can have the values bold|ital|boldital|roman|und|color.
     For now, emph (was emphasis) has been retained as an attribute for 
       most elements.

     Other:
   * End-tag omission was removed as an option for some elements (normally
       the TCIF SGML declaration will not allow end-tag omission anyway, but
       the option was retained where convenient, for nonstandard uses).
     Several attributes that had default values specified were made 
       impliable so they won't need to be explicit if SHORTTAG is set to NO.
     The attributes of Pt were changed from %a.rid; to most of %baseAtts; to 
       allow it to be used to mark up an alternative tree structure (used 
       with the remap attribute rather than an ID).
     .ex and .local entities were added to the definitions of several 
       entities (fdc, a.sect, a.olist, etc) for consistency, and the .core 
       intermediate entities were eliminated for simplicity.
     Long names of some entities were shortened or otherwise altered so 
       that the first 8 characters would be unique, in case a TIM document
       must be processed by a system that cannot handle long names.
     All attribute names were retyped in all-lower-case to be easily 
       distinguished from elements.
   ! The Table element was removed from pdc (the paragraph content model) as
       a simplification. This means that all tables must appear within 
       Frames (which are included in pdc).
     Comments within element declarations were made more consistent.

     Draft 2.5 differs from 2.4 (9/6/94) in the following ways:
     Some unnecessary entities have been eliminated (cdcEls, tdcEls, pdcEls,
       and sdcEls, which were defined in terms of other entities and used to
       define other entities). The entities cdc, tdc, pdc, and sdc have been
       redefined to approximately the meanings the ...Els entities had, and
       the content models of various elements have been adjusted 
       accordingly, e.g., by adding "(" and ")*" around the redefined "tdc".
     The FrameEls entity was renamed fdc for consistency.
   * The content models of TelPractice and Section1-Section5 have been 
       redefined to something that is equivalent in practice but easier to
       explain. For instance, the content of Section1 was defined as:
         ((Title|Title.comp)?,(Section2|%sdc;)*)
       but is now defined as:
         ((Title|Title.comp)?,(%sdc;)*,(Section2)*)
       Both mean that a 1st-level section has an optional Title followed by
       0 or more elements appropriate for a general structure (paras,
       frames, lists, alerts), followed by 0 or more 2nd-level sections. The
       old definition seems to allow for general-structure elements between
       or after Section2's, but a reader (or parser) would assume that any 
       elements appearing in those locations belong within a Section2. 
       Elements appearing before the first Section2, on the other hand, are 
       assumed to be introductory material for the 1st-level section. 
     The attributes pertaining to element marks (numbers, bullets, etc.) have
       been moved from the ATTLISTs to entities to allow easy redefinition.
     Entities for mark attributes have been added to the ATTLISTs for Frames,
       Alerts, Fnotes, and Listings). Mark attributes for these elements were 
       inadvertently omitted from the previous draft.
     The entity referencing the calstable.dtd has been renamed %table; to
       imply that another table model could be substituted.
     Declarations have been rearranged to simplify annotation and 
       documentation.
     Capitalization of element names has been regularized. All now begin 
       with a capital letter - this is done solely for readability, since
       names are converted to all-uppercase for processing, per tcif.dcl.
     Horizontal spacing has been regularized for readability.
     Many comments have been added.
   * Dates have been added to comments. -->
