news

harvest是一个下载html网页得机器人

This file describes recent user-visible changes in sgmls.  Bug fixes
are not described.  See the man pages for more details.

Version 1.2
===========

You can use catalog entry files in the format defined in the SGML Open
Draft Technical Resolution on Entity Management to map public
identifiers and entity names onto system identifiers.

Version 1.1
===========

The rast filter converts the output of sgmls to RAST format. See
rast.man.

There have been some changes to the output format: system and public
identifiers for external entities and notations are specified using
the new `s' and `p' commands; the filenames for external entities are
specified using the new `f' command rather than being arguments to the
`E' and `S' commands; the `\s' escape sequence is no longer generated;
an `S' command is generated for an entity mentioned in a `{' command;
RS characters are no longer stripped and are represented by `\012'.

The -c option now takes an argument giving the name of the file to
which the report of capacity calculations should be written.  The
report is in RACT format.

sgmls can now handle non-seekable devices (such as pipes).  If no
filenames are specified on the command line, sgmls will read the
document entity from standard input.  A filename of `-' can also be
used to refer to standard input.

The output format no longer distinguishes attributes whose declared
value is ID, IDREF or IDREFS and uses the TOKEN keyword for these.
(The ESIS does not distinguish these attributes.)

The -a option is no longer allowed.

Version 1.0
===========

Eight bit characters can be made name characters.

Ambiguous content models are always detected.  The -a option is ignored.

Version 0.8
===========

On Unix systems, the configure script can automatically generate
config.h from unix.cfg.

An APPINFO parameter in the SGML declaration other than NONE is output
as a `#' command.

sgmls has a -iNAME option which defines a parameter entity NAME with
replacement text "INCLUDE".

sgmls has a -u option that warns about undefined elements.

In SGML_PATH, the public text designating sequence (escape sequence)
is now accessed with %E rather than %S.  %S now gives the system
identifier.  When a system identifier is supplied, SGML_PATH will be
used to generate a filename only if SGML_PATH uses %S and the system
identifier does not contain (semi-)colons.

An accordance with Amendment 1 to ISO 8879:1986, in the SGML
declaration `ISO 8879:1986' rather than `ISO 8879-1986' is required.

With MS-DOS, errors can be redirected to a file using the -f option.

Version 0.7
===========

More quantity limits can be changed.  In particular, GRPCNT, GRPLVL
and GRPGTCNT can be increased up to 253; NAMELEN can be increased to
239.  Names are now dynamically allocated (that is, the amount of
memory used to store a name depends on the actual length of the name
rather than on value specified for NAMELEN.)

sgmlasp supports a -n option which disables upper-case substitution
for names in replacement files.

sgmls has a -p option which makes it parse only the prolog.

The format of the I command has been changed: CDATA and SDATA entities
are distinguished using a keyword rather than by surrounding SDATA
entities with \|s.

Version 0.6
===========

A concrete syntax can now be specified explicitly in the SGML
declaration, and various parts of the reference concrete syntax can be
changed, including some quantity limits.

The treatment of notation identifiers and the syntax of the N output
command have been changed.

SUBDOC support works with MS-DOS.

The -n option has been removed; references to subdocument entities are
always resolved.

It is no longer possible to change the implied SGML declaration using
environment variables.  You must use an explicit SGML declaration
instead.

The V output command has been removed.

sgmlsasp supports multiple replacement files.