Requirement: To facilitate automatic translation of well-defined document character sets with human inspectability of the translation information

(Note: This information is not required for SGML parsing.)

1. Add the ability to specify when the definition of a character set is represented using the format of the character set description parameter of the SGML declaration.
2. Add the ability, within a character set description, to identify a base set character by its string name (not just by its character number).
3. Add the ability to indicate when the string name for a base set character is taken from the ISO10646 character registry, rather than being the name used in the base set definition itself. (For ISO sets, the two are normally the same.)
4. Allow the base set of a formal character set definition to be itself a formal character set definition. This allows subsets of large character sets to be used in the definition of other character sets.

Requirement: To facilitate automatic generation of display character entity sets from definitional character entity sets.

For definitional character entity sets, add the ability to define the entity text as either a character in a defined character set and/or by an ISO10646 character registry name.

Requirement: To facilitate more powerful management of sophisticated versioning requirements.

Allow Boolean combinations of INCLUDE and IGNORE keywords in marked section declarations.

Requirement: To enhance the usefulness of capacities.

1. Make specification of any or all capacity limits in a document optional.
2. Make an SGML system's support for any or all capacity limits optional.
3. Add capacity limits for document instances, such as number of elements, number of data characters, etc.
4. Allow optional specification of actual capacities (not just capacity limits).

Requirement: To provide more flexibility for attribute design.

1. Allow a given name token to appear in the declared value of more than one attribute in the same attribute definition list.
2. Allow multiple attribute definition list declarations for a single element type.

Requirement: To simplify the specification of SGML declarations.

Allow reference to an existing SGML declaration with local modifications.

Requirement: To allow more choices among optional SGML features.

Modularize the SHORTTAG feature so that attribute minimization can be used with or without allowing empty tags.

Requirement: To clarify difficult portions of the text of the standard.

Clarify the description of record boundary handling, explaining clearly the relationship between detecting data characters and ignoring characters, including whether ignored characters are first recognized as data characters.

Requirement: To support multi-byte character sets with greater convenience

Devise a less burdensome method for declaring long sequences of character numbers.

Requirement: To facilitate name-space modularization in DTDs and LPDs

[The report did not include possible changes that address this requirement.]

Requirement: To support the requirements identified by an earlier review.

The recommendations in WG8 N1035 are considered to be incorporated into this list.

Requirement: To support multi-byte character sets with greater convenience

1. Provide a way to declare name characters that have no upper/lower case distinction without having to repeat the characters.
2. Provide a uniform character range specification mechanism and use it uniformly (e.g., in the UC/LCNMSTRT and SHORTREF specifications) to add large chunks of characters concisely).

   [WG8 N1854 addresses this requirement further.]

Requirement: To support multi-lingual text

Provide a means for specifying more than one language in the public text language of a formal public identifier.

Requirement: To enable SGML documents to be interchanged in a plain text format that is accessible by commonly used text editors and file utilities that are not inherently capable of interpreting SGML.

Revise the conformance clause to require that conforming SGML systems be able to use a plain text format in addition to any other format that they use.

Requirement: To integrate document architectures, formal system identifiers, and property sets and groves into SGML.

Incorporate the SGML Extended Facilities of the revised ISO/IEC 10744 into ISO 8879.

Requirement: To correct editorial errors

1. In clause 5.1 state that Figures 1 - 4 are based on the reference concrete syntax and that the character numbers used are those of the syntax-reference character set.
2. Remove references to ISO 2022.

Requirement: To support simple external references in SGML

It shall be possible, in an IDREF or IDREFS attribute value, to specify the name of the entity that defines the name space in which a referenced ID occurs. The syntax is "#ENTITY entity-name ID" (or possibly a list of IDs after which a suitable reserved word, such as #THISDOC, could be used to identify additional local IDs). This facility shall not be extended any further.

The standard will point out that HyTime should be used for more sophisticated requirements such as indirect referencing. The standard will also warn the user of the risks of using direct referencing to external objects whose location could change. (Note: The experience of one large user has been that any object referenced more than twice should be referenced indirectly.)

Requirement: To allow any SGML document to be a subdocument

SGML declarations will optionally be allowed in SUBDOC entities.

Requirement: To extend the character model of SGML to include entity and storage management.

The standard will be based on the character model that is summarized below. Terms used that are not defined here are defined in ISO 8879 or in the Formal System Identifier Definition Requirements (FSIDR) of the SGML Extended Facilities. Some terms defined in those standards are introduced here with informal definitions that are intended to be consistent with the formal ones.

NOTE: This text is subject to revision as the Review proceeds.

1. An SGML document is made up of entities, which are sequences of characters. Character sequences are passed to the SGML parser by the entity manager. As characters are abstract semantic constructs, and as computers can only deal with representations of abstractions, what actually passes to the parser is a non-standardized internal representation of the characters, known as the "system representation of characters" (SysRC).
2. The SGML declaration that governs a document defines a character set known as the "document character set" (DCS). A character set is the mapping of a character repertoire onto a "code set", which is an ordered set of consecutive bit combinations of equal size. A bit combination is an ordered collection of bits, interpretable as a binary number. The size of a bit combination in a code set (the "code set width") is the smallest power of 2 that is an integral multiple of 8 and which is such that 2, raised to that power, is at least 1 greater than the largest bit combination in the code set.
3. The DCS maps both significant characters and purely data characters to bit combinations in the code set. The base-10 integer equivalent of the bit combination mapped to a character is the character's "character number", which can be used in numeric character references in the prolog and document instances. The DCS identifies the remaining bit combinations of the code set as being mapped to "non-SGML characters"; they have character numbers. For each bit combination in the code set, even those that are non-SGML, there is a corresponding SysRC.
4. An entity is a virtual storage object. The entity manager actually maps an entity onto one or more real storage objects (or portions thereof). A storage object consists of "octet sequences" (or other formats) representing the stored characters, known as the "storage representation of characters" (StoRC). A StoRC is invalid if it does not represent an SGML character.
5. In the course of accessing storage objects, the entity manager, in conjunction with the storage manager(s), could invoke processes such as conversion, decompression, or decryption whose inverses might have been applied when the StoRC was stored, in addition to mapping from the StoRC to the character semantics (and therefore to the SysRC). Such processes and mappings could either be intrinsic to the storage manager or specified as an attribute in the storage object specification in the FSI.
6. The StoRC is mapped to the SysRC in one of two ways: either with or without the use of the document character set.
   • If the storage object is DCS-dependent a "bit combination transformation format" (BCTF) is applied to the StoRC to produce bit combinations, which are in turn converted to characters (and therefore to the SysRC) using the document character set.
   • Without the DCS, the storage object is associated with a mapping from the StoRC to characters (and therefore to the SysRC). This mapping, which could imply transformation of an octet sequence, is known as a "character repertoire encoding" (encoding).
7. DCS-dependent conversions are accomplished by the parser giving the entity manager the mapping from the bit combinations to the SysRC. The entity manager in turn passes it to the storage manager, which uses that to get from the StoRC to the SysRC.
8. Whether or not the DCS is used for mapping the StoRC to the SysRC, it always exists. Therefore, there is always a mapping from the DCS to the SysRC. That mapping is used to interpret numeric character references in the prolog and document instances, and also to determine whether a StoRC is invalid because it does not represent an SGML character.
9. The SGML declaration also defines a character set known as the "syntax-reference character set". Its purpose is to allow characters used in the defined concrete syntax to be referenced within the SGML declaration by character numbers that are not dependent on the DCS. It does not play a role in the mapping from the StoRC to the SysRC.
10. A "char" in a grove is an abstract data type representing a character (defined above as an abstract semantic construct). The concrete representation of a grove must allow any char to be distinguished from all other chars, and for the semantics of each char to be determined. The concrete representation may or may not rely on the DCS, UNICODE, and/or other character sets and repertoires.