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

---            Release of Lynx v2.8 (March 10, 1998)                ---

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

1998-03-10

* omit platform-specifics about release (recommended by HN). - TD

* drop samples/lynx.cfg since it is redundant (recommended by HN).  - TD

* modify makefile.in so that "make install-cfg" does not edit lynx.cfg

  to point at local helpfiles, while "make install-help" will now do

  that (reported by LP). - TD

1998-03-07

* move ifdef's for EXP_8BIT_TOUPPER inside UPPER8 function to allow

  check for DOS/WINDOWS display charsets. - LP

1998-03-06

* clarify relationship between the configure script and userdefs.h with

  respect to the LYNX_CFG_FILE definition. - TD

* newer CF_FIND_LIBRARY autoconf macro, to allow ncurses library in /lib

  only, needed for SUSE Linux. - TD

* add casts to UPPER8 function's comparison for equality of parameters, and

  to TOUPPER macro to guard against sign-extension due to prototype. - AAC

* correct spelling of -lcur_colr library in aclocal.m4, change order of

  tests to put -lHcurses second on HP-UX (patch by Jonathan Sergent).

* restore simple zmodem downloader script, displaced by recent undocumented

  change (there are some concerns about security or permissions with the

  alternate script).  - DK

* DOSPATH localfile patch:  corrected display of file://localhost/c:/,

  which was displayed as file://localhost/c%3A/ (tweak LYConvertToUrl() in

  LYUtils.c).  Now we may start "lynx c:" or 'go' c:  and not see these

  strange letters.  On the other hand, if % and # symbols found in local

  path it will be escaped as before (note that #fragment is not supposed here

  to be used from command line...) - LP

* workaround error in vendor's terminfo description of 'dtterm' by ensuring

  we limit color pair-number accordingly. - TD

1998-03-04

* strip -g compiler option if not otherwise specified, for debugging (note

  that autoconf 2.12 does not try to use -O option due to concerns about

  compiler optimization). - TD

* correct value of $host_os used in CF_CURSES_LIBS for special-case of

  HP's -lHcurses (Jonathan Sergent <sergent@ecn.purdue.edu>).

* correct def7_uni.tbl for Greek letters, note about UPPER8 assumption

  added to chrtrans/README.format. - LP

* few comments corrected. - LP

1998-02-28

* create new PRCS revision (2.8pre) for pre-release testing of Lynx 2.8 -TD

* remove RELEASE_STATEMENT, merging it into README as per HN's

  recommendation - TD

* comments/minor changes to HTDOS.c, HTDOS.h, LYUtils.c - LP

* document recent fix for RAW_DOS_KEY_HACK in lynx.cfg - DK

* rename CHANGES.new to docs/CHANGES2.8 - TD

* undo last minor change to UPPER8 - LP

1998-02-27

* add RELEASE_STATEMENT file. - TD

* modify slang configuration to add fake 'scrollok()' - JED

* modify LYmktime() so that 2-digit years must be greater/equal to 70 to be in

  1900's (patch by Bela Lubkin).  Also adjust parse_windows_nt_dir_entry() and

  parse_cms_dir_entry() as noted by Bela - TD

* comment-out entries in cp1252_uni.tbl which correspond to nonprinting

  control-characters - LP

* add logic in SGML.c to correspond with 2.7.2 logic for codes used by

  Microsoft FrontPage. - LP

* remove unneeded code for nbsp, ensp, emsp, thinsp and shy from handle_entity()

  in SGML.c - LP

* ifdef UPPER8 function with EXP_8BIT_TOUPPER,

  add configure option --enable-8bit-toupper to allow testing.  - TD

* case-insensitive search and 8bit letters:  It was realized that

  information about upper/lower mapping got from TOUPPER which depends on

  locale in its 8bit and usually fails on non-UNIX systems.  We introduce

  UPPER8 and strcasecomp8 to implement 8-bit aware case-independent search

  using one simple assumption:  8bit upper/lower case letters have their

  "7bit approximation" images in def7_uni.tbl matched case-insensitively

  (7bit).  The cost of this assumption is that several differently accented

  letters may be interpreted as equal but this is negligible if the search

  string is more than one character long.  (LYStrings.c, also HTString.c

  and LYMainLoop.c).  - LP, TD

* add SunOS 4.x-specific ifdef to quiet redefinition warnings for sys/ioctl.h

  vs sys/termios.h - TD

* convert usage/options messages into arrays to simplify maintenance. - TD

* add "-useragent" command-line option to set "User-Agent" (request by Merlin

  Mathesius <merlin_mathesius@swissbank.com>).  - LP

* tweak HTML in LYDownload.c, LYPrint.c to restore some double-spacing effects

  altered in 'collapsible spaces' change 1998-01-25 (reported by James Elkinton

  <zio@blueneptune.com>).  - LP

* use strncasecomp/strcasecomp consistently throughout to avoid library

  dependency on strncasecmp/strcasecmp. - TD

* correct lengths in several strncasecmp calls for command-line options

  (though -w is still an abbreviation for -width). - TD

* test-build with BSD curses on OS/2 EMX (does not display properly), and

  with ncurses -TD

* add configure check CF_FUNC_LSTAT -TD

* modify CF_CHECK_ERRNO so that configure script on OS/2 EMX finds

  sys_errlist[] -TD

* reset styles[] and styleSheet list in DefaultStyle.c, HTML.c so that when

  a user reloads a page with ^R, we undo changes to those data structures

  (reported by Larry Virden). - TD

* remove extra copy of HTML.h from Implementation directory. - TD

* add CF_GCC_ATTRIBUTES to configure script, to use in flushing out unused

  parameter warnings (and later, to support prototypes for logging). - TD

* rename 'extra_entities' to 'unicode_entities' (recommended by LP) - TD

* add configure option --use-default-colors, to allow background color to use

  default value (request by SUNAGAWA Keiki <Keiki_Sunagawa@yokogawa.co.jp>) -TD

* change default for --disable-internal-links so it is normally-off - TD

1998-02-19

* add configure test CF_FIONBIO, some ifdef's (e.g., S_IFLNK) to support

  build on OS/2 EMX. - TD

* add configure option --disable-extended-dtd, for testing - TD

* split-out entities.h from HTMLDTD.c - TD

* remove obsolete mapping in "private" e000 area of mnem_suni.tbl,

  mnem2_suni.tbl, rfc_suni.tbl - LP

* remove unused function HTMLGetLatinOneValue() - LP

* remove logic that would attempt lookup of Unicode entities by index into the

  unicode_entities.put_entity method (e.g., of HTML_dtd), in HTML.c

  (HTML_put_entity) and HTMLGen.c (HTMLGen_put_entity) - LP

* tweaks to HTPlain.c, LYCharUtils.c and SGML.c, remove extra checks for

  Unicode mdash, ndash, trade, as well as "iso-8859-2", "koi8-r", since they

  are done in the chartrans tables - LP

* add an alias 'H' for F1 when using RAWDOSKEYHACK (LYStrings.c) - DK

* restore -DRAWDOSKEYHACK to src/makefile.dos, reported by DK - TD

* correct a typo in CF_CURSES_LIBS configure macro, which caused the configure

  script to not add termcap library to the list - TD

* correct --disable-internal-links, by adding DONT_TRACK_INTERNAL_LINKS to

  config.hin - TD

1998-02-13

* use size_t in SGML_dtd. - TD

* remove extra trademark symbol and quotes from chrtrans/*.tbl, since

  they are in the default def7_uni.tbl - LP

* New item to lynx.cfg added:  PREPEND_CHARSET_TO_SOURCE (similar to BASE), we

  need it to resolve charset for downloaded/printed HTML sources which became

  local.  This is really useful because downloaded and printed texts may

  happened to have different charsets so assume_local_charset fails.  For

  downloaded source meta charset added from HTTP charset header (if and only if

  present).  For printed source meta charset added from display_charset (real

  charset after chrtrans).  If the original source has its own meta charset

  and we add the second on top - Lynx currently use those on top and ignore

  others (old) value (that is correct, although non standard like BASE on top).

  The compilation default is FALSE (compatibility).  - LP

* modify ncurses mouse behavior to match slang (use button 3 rather than

  button 2 for "previous document" (reported by Michael Ritzert) - TD

* use case-independent comparison for tail-only comparison in override_proxy

  (patch by pg@sweng.stortek.com) - TD

* remove unneeded definitions of ALT_CHAR_SET, NCURSES_VERSION in

  src/makefile.dos - DK

* correct syntax for the RAW_DOS_KEY_HACK in lynx.cfg, add a little information

  on how to compile this.  Also patched makefile.dos to make compilation of the

  raw dos key hack the default when compiling in DOS. - DK

* change back to boolean since the logic was unused: UCCanTranslateFromTo(),

  UCCanTranslateUniTo(), and UCCanUniTranslateFrom(). - TD

* Mods of SGML.c and LYCharUtils.c to improve handling of entities,

  obsolete stuff removed (not all yet). - LP

* -localhost mode: disable interpreting BASE HREF= if local_host_only

  to resolve local html files with BASE (Lynx add base to downloaded

  sources and this may affect links if looking them with -localhost).

  (tweak in html.c). - LP

* Modify HTMLGetEntityUCValue in LYCharSets.c:

  we begin to move from old style entities[] to unicode-based. - LP

* added a few comments. - LP

1998-02-07

* Fixed a bug in HTTP.c which could cause trashing of authorization info

  for an origin HTTP(s) server if the request is being proxied. - FM

* Tweak in HTTP.c so that the reloading variable is not reset until

  after any resubmissions due to authorization challenges. - FM

* Tweaks of userdefs.h, lynx.cfg, LYMessages_en.h, LYMain.c, LYMainLoop.c,

  LYClean.c, LYReadCFG.c, and LYGlobalDefs.h so that the QUIT_DEFAULT_YES

  compilation symbol can be overridden in lynx.cfg, in keeping with

  the principle that only strict security-related compilation symbols

  cannot be overridden at run time. - FM

1998-02-05

* adjust formatting, comments & some messages in WWW files to simplify

  comparison with 2.7.2 version.  (I have preserved some chunks of difference,

  ifdef'd NOTUSED_FOTEMODS or NOTDEFINED for ease of comparison -- these may be

  removed at a later time, when we are done absorbing changes from 2.7.2) - TD

* Update SGML entity to Unicode table in HTMLDTD.c from ftp.unicode.org -

  lots of new entities, iso-latin-1 now included there.

  See also new files in test/ directory:  sgml.html and unicode.html.

  Tweak a few typos in test/*.html according to sgml.html.

  Add a few new entries in def7_uni.tbl according to sgml.html - LP

* Cleanup iso01-iso09, cp866 tables against ones found at ftp.unicode.org -LP

* Strip the fat from LYCharSets.c - unnecessary obsoleted tables removed

  and redirected to SevenBitApproximation table (ones which have unicode tables

  override it anyway, those for CJK had exactly the same tables as

  SevenBit..). - LP

* Remove obsolete "Other ISO Latin" charset, which works _exactly_ the same

  as "7 bit approximation" (See the note on 06-14-96 where it comes from) -LP

* Sort the list of charsets in Options according to 2.7.2 for 1997-10-15,

  see the comments in UCinit at the bottom of UCDomap.c - LP

* add ifdef to HTFile.c to workaround conflicting S_IFIFO vs S_IFSOCK

  (e.g., on Apollo, from a report by Dave Eaton <dwe@arde.com>) - TD

* updated INSTALLATION (patch by HN) - TD

* modify DOS handling of CTL-C and CTL-BREAK handling for DOS to allow

  exiting via SIGINT when pressing CTL-C or CTL-BREAK regardless of the

  BREAK setting in DOS.  This leaves the BREAK setting unchanged unless

  CTL-C or CTL-BREAK is actually pressed (in which case it leaves BREAK

  ON) - DK

* add note about Lynx vs frames to User's Guide (from Al Gilman) - TD

1998-01-27

* correct lynx.man to reference section "1" (one), not "l", since it is

  not installed in the latter (reported by Jason Castonguay

  <jcast@ntplx.net>) - TD

* fix for VMS declarations of mainloop() and from_hex() (reported by FM) - TD

* add to include-paths in libmake.com, build.com (reported by FM, citing

  Brian Tillman) - TD

* revert HTML.c to use LYHandleSELECT() - TD

* align cases for function keys in LYStrings.c with 2.7.2 - TD

* remove EXP_CHARTRANS ifdef's - TD

* remove symbol USE_SIZECHANGEHACK -TD

* adjust formatting, comments & some messages in most remaining ./src files to

  simplify comparison with 2.7.2 version.  - TD

1998-01-25

* Changed the utf-8 Display Character Set string to "UNICODE UTF-8"

  (i.e., added the hyphen between UTF and 8), and fixed typos in

  the comments about iso-8859-5. -LP, FM

* Modified HTML.c to add collapsible spaces before and after IMG ALT

  strings or pseudo-ALTs whether or not the IMG element is within

  Anchor content.  The occasional absence of such spaces due to poor

  HTML was problematic for blind users.  Note that markup which uses

  an image for a fancy letter with the ASCII letter as an ALT will

  have that letter separated from the rest of the word.  Also note

  that these spaces will not be collapsed in PRE blocks, but IMG is

  invalid in PRE blocks.  The spaces could be omitted when IMG is

  used invalidly in PRE blocks, but that would revive the problem

  for blind users that they might not be present at all, so it seems

  better to suffer extra spaces in such cases. - FM

1998-01-24

* Tweak in HTAccess.c to reset the title element in anchors on forced

  reloads. - FM

1998-01-22

* Tweaks in HTML.c to prevent collapsing of blank lines within TEXTAREA

  default values.  Note that we still strip any leading or trailing

  blank lines (reported by Edward S. Marshall <emarshal@xnet.com>). - FM

* Tweaks of LYCurses.c, LYMain.c, LYOptions.c, LYReadCFG.c, LYUtils.c

  and LYAuto.c to ensure properly setting, with internal buffering, and

  avoid memory leaks for the display and homedir values (reported by

  pg@sweng.stortek.com). - FM

* include userdefs.h in LYCurses.h to address redefinition warnings for

  curses-related configuration. - TD

* adjust formatting, comments & some messages in recently-modified files to

  simplify comparison with 2.7.2 version.  - TD

* remove obsolete logic from UCCanTranslateFromTo, (recommended by LP). - TD

* make character set names (e.g., cp437, cp850, next) consistent in lynx.cfg,

  userdefs.h (reported by LP).  - TD

* change configure script to check for termcap library before curses, in case

  it is dependent.  - TD

* More corrections in the Russian Cyrillic area of def7_uni.tbl. - LP

1998-01-17

* correct logic for no_color_video by ensuring that if not defined, it

  doesn't disable underline with color. - TD

* remove NSL_CLONE - didn't build. - TD

1998-01-16

* eliminate gcc "shadowed variable", "nested declaration", "no previous

  prototype" warnings. - TD

* add --disable-internal-links configure option. - TD

* remove W3C gateway (patch by DK, based on email from Gerald Oskoboiny

  <gerald@w3.org>). - TD

* documentation updates (patch by JS). - TD

* add CONTENT_TYPE to environment of lynxcgi POST documents (reported by

  Ty Sarna <tsarna@endicor.com> & FM). - TD

* add NSL_CLONE ifdef that enables use of the Linux clone() call instead of

  fork() for the gethostbyname() calls in HTTCP.c.  (patch by Ryan Nielsen

  <ran@haxor.com>). - TD

from the lynx code.  I think the following patch does this.

1998-01-15

* Tweak of the NSL_FORK code in HTTCP.c to prevent zombie processes from

  being created when gethostbyname() is interrupted. - RN

* Corrected the Russian Cyrillic area of the def7_uni.tbl Unicode to

  7-bit approximations conversion table. - LP

* Mods of UCAux.c, LYCharUtils.c, SGML.c and HTPlain.c to use 7-bit

  approximations instead of 8th bit stripping when the document charset

  is koi8-r and the Display Character Set does not have the corresponding

  characters. - FM

* Tweaks in HTML.c to ignore a newline character immediately following

  a PRE start tag more consistently.  When PRE blocks are used to create

  blank lines, you should get only as many as are actually included in

  the PRE block, unless you include less than the number specified for

  "before" the block to which you return, in which case you'll get the

  latter number of blank lines. - FM

1998-01-13

* Fixed a typo in the HTParse() code of HTParse.c for checking whether

  a colon in the host field indeed delimits a port number. - FM

* Changed the internal MIME names for all Microsoft Windows codepages

  from cp#### to windows-####, since the latter is used more commonly

  as the charset value in META elements and/or Content-Type headers,

  and most of them have been registered that way with IANA.  Note,

  however, that Microsoft uses cp instead of windows- prefixes in the

  charset names for the Unicode conversion tables submitted to the

  Unicode Consortium.  Also note that we still treat cp, cp-, and ibm

  prefixes as synonyms for the windows- prefix. - FM

1998-01-06

* work around for 'keypad()' function w/o corresponding key defines

  (reported by Francisco A. Tomei Torres for NeXTstep 4.0). - TD

* quote message in install rule, just in case (from HN). - TD

* correct expression to supply default value of .cfg file's pathname

  (reported by LE & Ambrose Li <acli@acli.interlog.com>). - TD

* correct dependency of .cfg files upon $(libdir) in makefile.in - TD

1998-01-04

* correct temporary filename for -traversal -crawl (patch by Doug Kaufman). - TD

* correct redefinition of SYSTEM_MAIL_FLAGS. (reported by Doug Kaufman). - TD

* uncache current document when character set may have changed (patch

  by Andrey A. Chernov) - TD

* modify Linux setfont code to restore the original font on exit - TD

* correct missing backslash (again) in makefile.in

  (reported by Mark Mentovai - again) - TD

1998-01-02

* workaround for sed expression when rerunning configure with config.cache

  for slang libraries (reported by Hynek Med). - TD

* fix indicated by Laura Eaves (LE) for LYUCFullyTranslateString_1() - TD

* modify SPAWNING_MSG for DOSPATH configuration, avoid unnecessary use

  of clear-screen in shell (patch by Doug Kaufman). - TD

* remove install-help from list for install rule, adding message to

  indicate how to install it. - TD

* install lynx.lss only if color-style is configured. - TD

* correct filemode for DJGPP when processing config-file, patch by

  Doug Kaufman). - TD

* add LYNX_LSS_FILE to config.hin, reported by Ambrose Li

  <acli@acli.interlog.com>. - TD

* new copy of INSTALLATION file from HN. - TD

* add COPY* to install rule. - TD

* simplify includes in UCdomap.c by using include-path in makefiles. - TD

* build configure script with patched autoconf to workaround limited

  environment space on CLIX 3.1 - TD

* rename mnemonic_suni.tbl to mnem2_suni.tbl to accommodate 'make' on

  older system with 14-character name-limit. - TD

* sorted some lists (e.g., character sets) which were unreadable. - TD

* incorporate most of FM's changes from 971124 to 971230 (see below). - TD

  (Note:  don't try to use fflush(stdin) -- it's an FAQ -- nonstandard,

  dumps core on some platforms).

1997-12-30

* Added Unicode translation support for "DosBaltRim (cp775)" and

  "WinBaltRim (cp1257)". - FM

1997-12-29

* Fixed bug in the order of re-initializations for the 1997-12-22

  get_connection() mods. - FM

* Fixed typos in some LYCharUtils.c comment. - FM

* Fixed typo in LYrcFile.c comments concerning the NO_FROM_HEADER

  symbol. - DK

1997-12-24

* resolve conflict between autoconf $INSTALL and lynx's INSTALL_PATH

  (reported by Scott McGee). - TD

* modify CF_CHECK_FUNCDECL to work with non-gcc compilers. - TD

* add <stdio.h> to prototype test for getgrgid getgrnam, needed on SVr3. - TD

* add install-lss to makefile.in. - TD

* modify order of includes in LYKeymap.c so that USE_EXTERNALS is defined

  if it comes from userdefs.h (inferred from a report by Larry Virden). - TD

* correct spelling of USE_EXTERNALS which prevented --enable-externs from

  working. - TD

* move assignment to stderr from VMSexit to cleanup function (reported

  by Brian Tillman & FM). - TD

* modify UCChangeTerminalCodepage() to not use stderr, disentangling it

  from stderr/logfile hack. - TD

* rename docs/CHANGES2-* files to docs/CHANGES2.* (reported by

  Doug Kaufman). - TD

* add start_color() call to DOSPATH version of start_curses() to

  fix problem reported by Doug Kaufman - TD

* correct typo in ColorCode() function (reported by Doug Kaufman) - TD

* modify comments in makefile.in (patch by HN) - TD

* minor fixes for CF_CHECK_ERRNO, CF_SIZECHANGE, CF_SLANG_CPPFLAGS

  macros. - TD

* add configure option --enable-nsl-fork to turn on code defined by

  NSL_FORK. - TD

* add configure option --enable-links to turn on code defined by

  UNDERLINE_LINKS. - TD

* add configure option --enable-font-switch, to turn on code defined by

  EXP_CHARTRANS_AUTOSWITCH. - TD

* modify configure CF_CURSES_LIBS test to look also for ncurses,

  since some systems do not have a "curses" library (reported by

  Michael Ritzert). - TD

1997-12-22

* Reworked -width=NUMBER command line switch for specifying the screen

  width for formatting of dumps (default is 80; forced wraps occur at one

  column less than the screen width, i.e., at 79 by default). - FM

* Tweak of get_connection() in HTFTP.c to re-initialize the elements of

  the connection structure when using a previously allocated connection

  object. -DW, FM

* Mods of LYexit.c, LYClean.c, LYMain.c and LYMainLoop.c to fflush() all

  standard file descriptors in conjunction with all fopen() and fclose()

  operations on the TRACE log and redirections of stderr via equality

  statements (see 1997-12-16, 1997-12-13 and 1997-07-18 mods). - FM

1997-12-18

* forgotten last } UN-inserted (John Line) - WB

* few commented messages corrected according to RFC2068 (HTTP1.1)

  (Leonid Pauzner) - WB

1997-12-17

* Changed all %i tokens to %d or %ld in functions that always get decimal

  values in their arguments. - FM

1997-12-17 (all patches by Leonid Pauzner)

* open text mode while writing lynx.rc and bookmarks files

  LYrcFile.c and LYBookmark.c (forgotten last } inserted). - WB

* add a comment about binary mode in LYMain.c and pile them

  in one ifdef - WB

* few misprinted characters (in comments) corrected in

  \chrtrans\cp850_uni.tbl - WB

1997-12-16

* Further mods in LYClean.c (see 1997-12-13 mods) do block the

  *stderr = LYOrigStderr;  statement in cleanup() for VMS with VAXC. - FM

* Added an ELGOTO keystroke command, mapped by default to upper case 'E'

  (lower case 'e' still invokes the editor for the current document, if

  it is a local file), for passing the URL (or ACTION) of the current

  link to the line editor for editing as a goto URL.  Works like the

  ECGOTO ('G') keystroke command for editing the URL of the current

  document as a goto URL, i.e., if the URL is not modified, is set to

  a zero-length string, or ^G is entered, the command is cancelled. - FM

* Mods of LYMainLoop.c to disallow ECGOTO or ELGOTO commands for File

  Management (DIRED_SUPPORT) menu URLs. - FM

1997-12-15

* Added reporting of the Discard value (YES or NO), in the Cookie Jar Page

  (note that until we have code for storing cookies beyond the current

  session, Lynx behaves as if Discard were set, whether or not it was set

  by the server, but we still report the server's instructions, as we do

  for the Maximum Gobble Date). - FM

1997-12-14

* added a lynx.cfg switch to turn on/off extended dos translations for keys.

  -WB

* added minor DOS tweaks by Doug Kaufman and myself. - WB

* fixed windows binary mode piping. - WB

1997-12-13

* Tweaks in LYList.c to make the link numbers correspond with those in

  the current document when LINKS_AND_FORM_FIELDS_ARE_NUMBERED in on,

  and to show the current document's URL in the display. - FM

* Mods in HTTP.c to show the body of a 305 return status instead of ever

  acting on it.  A number of security/privacy problems with this status

  are being discussed in the HTTP-WG, and it's likely to be dropped or

  deprecated in the next revision of the HTTP/1.1 specs. - FM

* Restored the v2.7.1 HTmmdecode() that's specific for iso-2022-jp in

  HTMIME.c.  We still only call it when HTCJK == JAPANESE, and the

  generalized version reportedly has problems. - FM

* Removed the *stderr = LYOrigStderr; statement from VMSexit() in LYCurses.c

  and added it to the if() claused section which closes the TRACE log file,

  if opened, in cleanup() of LYClean.c.  This is just a stab in the dark

  for dealing with reported problems on exit with VAXC on VAXen.  We may

  need to use freopen() for redirecting TRACE messages to a file for VAXC

  or VAXen (there's no apparent problem with the current code with DECC

  on VMS/AXP). - FM

* Added more function key cases to the keypad() mode switch() in LYgetch()

  of LYStrings.c.  We don't need them with this code set, but they may as

  well be there for completeness. - FM

* check for and use ncurses default-color extension, use to make the

  color-style configuration usable on xterm with white background. - TD

* add cases in LYgetch() for remaining VT220-compatible codes, reported

  by Michael Ritzert and FM.  - TD

* add configure test to see if slang can be linked without curses or

  termcap libraries, request by John Davis. - TD

* set umask in mkdirs.sh script to work around users who have set root's

  umask to 077. - TD

* split up install rule to install-bin, install-cfg, install-man and

  install-help. - TD

* suppress suffixes in top-level makefile so 'make' won't try to create

  a file named "install" from install.sh - TD

* make $(CPPFLAGS) inherit consistently from top-level makefile. - TD

  (reported by Mark Mentovai)

* add configure option --enable-externs, to turn on USE_EXTERNS. - TD

* use explicit comparison for enum values, fix compile problem with VAX C

  for UCAux functions (reported by Brian Tillman). - TD

1997-12-03

* Remove CHANGES.configure, since it is redundant. - TD

* Make -g3 option check for OSF/1 3.2 consistent with -Olimit check. - TD

* Use 'result' rather than 'time' in mktime.c to avoid shadowing. - TD

* Use wborder only if configure finds it, not ifdef FANCY_CURSES. - PHDM

* List of systems having strerror replaced by configure test. - PHDM

* Beautifying of mktime.c (`now' was misleading). - PHDM

1997-12-02

* Mods in LYCookie.c to prompt whether to accept (or reject) cookies with

  an invalid domain value which yields a hostname prefix with a dot (as

  specified in section 4.3.2 of the IETF cookie drafts and RFC), rather

  than rejecting it silently.  Note that it is rejected in non-interactive

  (-dump or -source) mode, because the user can't be prompted. - FM

1997-11-27

* Added a PREPEND_BASE_TO_SOURCE configuration symbol in lynx.cfg which

  can be set FALSE to suppress prepending of a Request URL comment and

  BASE element to text/html sources retrieved for 'd'ownloading or passed

  to 'p'rint functions.  Note that this prepending is not done for -source

  dumps unless the -base switch also was included on the command line, and

  the latter switch overrides the setting of the PREPEND_BASE_TO_SOURCE

  configuration variable. - FM

1997-11-25

* Mods in LYMail.c and LYPrint.c to handle to=address(es), cc=address(es),

  subject=string, keywords=word_list, and body=message fields in the

  ?searchpart of mailto URLs as specified in the 1997-11-16 draft:

    ftp://ftp.isi.edu/internet-drafts/draft-hoffman-mailto-url-03.txt

  The cc=address(es) values are now handled as a Cc: header, and a self

  copy, if confirmed on prompt, is added to the Cc: header.  Note that

  the draft now allows mailto:?to=address(es)[&foo=blah] as a format,

  i.e., without an address before the ?searchpart.  The Netscape

  ?searchpart tack-on broke all pre-existing implementations of mailto,

  and this breaks Lynx's support through v2.7.1.  Though it's now

  supported in this code set, don't use it.  Complaints and formal

  requests to provide this functionality via a new scheme (e.g.,

  mail: or smtp:) without breaking existing implementation of mailto:,

  including from TimBL (HimSelf :), thus far have fallen on deaf ears.

  You can send email to those deaf ears via:

  	mailto:masinter@parc.xerox.com%20(Larry%20Masinter)

  Don't allow the IETF backward compatibility principle to go down

  the drain. - FM

* Mods to support use of the PMDF SEND command on VMS in conjunction with

  a temporary header file, so that all headers specified in the ?searchpart

  of mailto URLs can be included in the mailing, and made that the compilation

  default in userdefs.h to go with the default MAIL_ADRS mask for PMDF (If

  the "generic" VMS MAIL utility is used, not all of the headers can be

  specified on the command line, as it requires).  Look at the code in

  LYMail.c and LYPrint.c that is used when isPMDF is TRUE as a model of how

  to do this with other header-file-capable VMS mailers (but I don't have

  any besides PMDF, and can't do it for you :). - FM

* Tweak of the 1997-11-17 URL parsing mods so that relative symbolic

  elements immediately preceding a ?searchpart are handled identically

  to when a ?searchpart is not present. - FM

* Added support for setting both SYSTEM_MAIL and SYSTEM_MAIL_FLAGS in

  both userdefs.h and lynx.cfg. - FM

1997-11-23

* integrate PHDM's patch. - TD

* Check for existence of sys/wait.h. - PHDM

* Do not include sys/stat.h twice. - PHDM

* New autoconf-macros to check if a function is declared after including a

  specific set of files. - PHDM

* Declare strstr if it is not declared by string.h. - PHDM

* Declare getgrgid and getgrnam if not declared by grp.h. - PHDM

* Provide strstr and mktime implementations, and use them if needed. - PHDM

* Always include sys/types.h in HTUtils.h. - PHDM

* define NO_GROUPS if we do not have getgroups. - PHDM

* define NEED_REMOVE also if test yields unknown. - PHDM

* define mode_t if not done by system include files. - PHDM (phdm@macqel.be)

* change CF_UNION_WAIT to use compile+link rather than compile, to ensure

  that we pick up the wait-macros rather than an implied function, which

  should fix configure problems on NeXT for Francisco A. Tomei Torres

  <fatomei2@mail.unm.edu>. - TD

* spelling nit from Larry Virden. - TD

* add osf3.0 to special case compiler options using -g3 in place of

  -O -g, from Harsh Kumar <harsh@krcl.ernet.in>. - TD

* remove redundant "-c" option from $(installbin) in makefile.in, reported

  by Michael Warner <warner@wsu.edu>, for OSF/1 4.0. - TD

* if configure --srcdir is not used, comment-out rmdir's in "make distclean"

  rule. - TD

* modify recent erase/clear change so that it applies only to FANCY_CURSES,

  since BSD curses does not otherwise clear the screen when restoring after

  popup. - TD

* modify change for boxes so this builds with BSD curses, which has no

  wborder() function. - TD

* reverse ifdef for NO_SIZECHANGEHACK to USE_SIZECHANGEHACK, which is

  normally off, since this code did not work with BSD 4.4 curses, SVr4

  curses, or ncurses 4.1 with xterm. (Will make a configure test if anyone

  can identify a configuration that works). - TD

1997-11-21

* Changed LYE_LINKN to LYE_LKCMD (still mapped by default to Ctrl-V; may

  need to be entered twice for some systems) for invoking a prompt to

  enter any Lynx keystroke command when in form text fields. - FM

1997-11-20

* Added an LYK_CLEAR_AUTH command, mapped to underscore ('_') by default,

  which clears all authorization info for the current session.  This is

  recommended for the situation when one leaves one's terminal but doesn't

  want to end the browser session, nor want others to be able to retrieve

  protected documents with previously entered username and password info.

  Note that we don't yet have a way to block access to any protected

  documents which are still in cache. - FM

1997-11-19

* Tweak of HTTP.c to ensure closing of the connection when interrupted

  ('z'apped) on the first socket read. - FM

1997-11-18

* Mods to use the explicit label "Bad HTML:" in trace messages corresponding

  to BAD_HTML_USE_TRACE messages. - FM, KW

* Tweak of the restored TABLE-in-PRE block handling to deal with the

  case when it's encased in a CENTER or DIV with ALIGN="center". - FM

* Changed docs/README.defines to refer to INSTALLATION file - HN

1997-11-17

* Changed the default ASCII BOXVERT and BOXHORI definitions from '*' to

  to '|' and '-', and new function LYbox() for non-VMS (n)curses. - FM

* Use explicitly specified characters for drawing the corners of popup

  boxes with non-VMS (n)curses if we have also specified characters for

  vertical and horizontal lines. - KW

* Mods in HTTP.c to trim any "LYNXIMGMAP:" leader from Referer header

  values and strip the #fragment when making http(s) requests via links

  in the pseudo-documents constructed for MAP elements. - FM, KW

* Mods of SGML parsing to better handle SELECT blocks with invalid markup,

  in effect for both TagSoup and SortaSGML parsing (for the latter just

  in case it doesn't already prevent invalid markup from being passed on).

  Only OPTION elements are accepted in SELECT blocks.

  Otherwise, if the element is a form-related start tag, the

  SELECT block is closed and the new form-related element is started.

  All other elements are ignored.  A FORM end or start tag ends any open

  SELECT block and the current FORM (before starting another FORM if

  it is a start tag).  Note that we now support SELECT blocks outside

  of FORM blocks, which some providers are using for "informational"

  popups, but no Lynx user would do anything that awful, right? - FM

* Tweak in LYMainLoop.c to override an LYresubmit_posts flag if we are

  reloading a popped document due to a NOT_FOUND or NULLFILE return

  value from getfile(). - FM

  (But if the document that would be loaded after such a failure has POST

  data and is not in memory, still jump to the PREV_DOC code, see enttry

  for 1997-11-03. - KW)

* Worked in a variant of AJL's suggestion of setting LINES and COLUMNS

  artificially high before the initscr() call with Unix curses, then

  using a dummy call to size_change() for setting LYlines and LYcols

  to the actual screen dimensions, so that increases in the screen size

  during the Lynx session, and not just decreases, can be handled.

  Note that one still needs to do a RELOAD (^R) overtly to create a

  new HText structure based on the current screen size. - FM

  (The new code can be disabled by defining NO_SIZECHANGEHACK, in case

  some curses version or terminal cannot deal with it. - KW)

* Tweaks of the chartrans code. - FM, KW

* Change in HTNews.c to ignore very long bogus newsgroup names when listing

  available groups. - KW

* Tweaks in code for highlighting search text in anchors, to better deal with

  terminals in UTF-8 mode. - KW

* Mods to ignore anything past a '?' character when simplifying URL paths,

  as is now more clearly specified in the Fielding URL draft.  Note that

  we still assume a ';' is not in fact a parameter delimiter, or won't

  have a value with characters that could be misinterpreted as relative

  path specs. - FM

* Increased the redirection limit to the original Lynx's value of 10,

  because the HTTP/1.1 draft no longer will restrict it to 5, but instead

  will leave it to the discretion of the browser (and I like 10, as a

  reminder of the "good old days" :). - FM

1997-11-15

* Mods of parsing in HTML.c to restore support for the TABLE-in-PRE-block

  HTML kludge described in http://ppewww.ph.gla.ac.uk/~flavell/tablejob.html,

  which stopped working in conjunction with the 1997-05-21 mods (see below)

  for treating TABLEs as DIVisions to support ALIGNment attributes in TABLE

  markup.  The PRE-in-TABLE-block HTML kludge was not affected by those mods

  but works only with the TagSoup parsing mode. - FM, KW

1997-11-12

* Made handling of the 'x' in hexadecimal character references (&#x;;)

  case insensitive, as for the "alphabetic digits" (a-f, A-F). - FM

* Added a LINKN command key in the line editor, mapped to Ctrl-V, for

  invoking the F_LINK_NUM prompt when in form text fields.  Ctrl-V is

  still a dead key under other circumstances in the line editor. - FM

* remove options --with-lss-file and --with-cfg-file, since the corresponding

  files will always be installed under $(libdir), reported by Henry Eric

  Nelson (HN).  - TD

* modify configure script and makefile templates to permit builds with the

  --srcdir option, noted by Philippe De Muyter. - TD

1997-11-08

* Added Unicode support for "DosArabic (cp864)", "WinArabic (cp1256)",

  "DosGreek (cp737)", "DosGreek2 (cp869)", "WinGreek (cp1253)",

  "DosHebrew (cp862)", and "WinHebrew (cp1255)".  Note that we still

  have no code for taking DIR attributes, or directionality in general,

  into account. - FM

* Mods to handle "cp-#" charset names as synonyms for "cp#", and some

  cp# names as synonyms for "windows-#". -FM, KW

1997-11-07

* Change of LY_SLrefresh() in LYCurses.c as suggested by JED - FM

* add logic to make install-directories as needed. - TD

* remove --disable-char-trans option (ifdef's still remain). - TD

* correct type used in first compile-test for union wait. - TD

* correct order of deletions in "make distclean" rule. - TD

* new versions of macros used to check for ncurses version. - TD

* don't suppress -I/usr/local/include when checking -I/usr/local/include/*,

  to accommodate older versions of gcc. - TD

1997-11-06

* Use erase() instead of clear() in some places, to avoid unnecessary full

  screen repainting even for UNIX (and similar) curses.  Renamed LY_SLclear

  to LY_SLerase, but it is used for erase() and clear() now, so there should

  be no change in behavior for slang introduced by this. - KW

* Added new format for lines in chartrans .tbl files: U+NNNN "a string" where

  the string can have C backslash-escaped characters.  This allows to write

  replacement strings with 8-bit characters that won't be garbled when the

  .tbl files are transferred in some non-binary way.  Note there is no ':'

  between the Unicode value and the beginning quote character. - KW

* Changes to HTMLDTD.c (for SortaSGML parsing) and HTML.c (second line of

  defense) for dealing with invalid SELECT tags.  Also tolerate SELECT

  outside of forms. - KW

* Change in HTLocalName for DOSPATH so that it doesn't depend on whether

  HTDOS_name returns a pointer to static memory or not. - KW

* Started new file docs/README.defines.  Renamed CHANGES to CHANGES2-7,

  removed duplicate 2-7-1 entries.  Minor changes to INSTALLATION, other

  files.  CHANGES.chartrans is gone. - KW

* Tweaked color style stuff again.  Statusline may behave better. - KW

* Don't use memmove for HEAD in HTNews.c. - KW

* For UNIX, use normal umask permissions, instead of the extra-cautious temp

  file permissions, for files saved to disk by the default actions for PRINT

  and DOWNLOAD. - KW

* Changes (untested) in the code dealing with reading of compressed files in

  HTFile.c, in parallel to the code for other systems. - KW

* Mods to avoid va_start redefinition warnings when building with slang. - FM

1997-11-05

* Change in LYForms.c to avoid full screen repaint after popup is retracted

  for slang. - KW, FM

1997-11-03

* Internal links code revised.  Some details follow, but first a summary

  of the basic premises.  Links in a document given as href="#fragment" or

  href="" (empty, but href attribute is present) mean something different

  from href="url#fragment" or href="url" (where url is a non-empty absolute

  or relative URL not including a fragment), they are internal links pointing

  to an element or a location within the same copy of the document (or to the

  document itself, for href="").  In contrast, a URL-Reference with a non-

  empty URL refers to a (generally...) external resource.  If the "url"

  happens to coincide with (or, if it is relative, resolve to) the URL from

  which the current document was retrieved, the current document is a

  representation of the resource, but not the same thing.  In most such cases

  we can treat href="#fragment" and href="url#fragment" the same way, i.e.

  activating such a link is interpreted as a request to reposition within

  the current document; we are using the current document as a cache copy

  of the resource (like we would other documents in the memory cache).  But

  if we have marked our copy as no-cache, it seems more consistent to treat

  activation of a href="url#fragment" link as a request for retrieving a

  new copy and _then_ locating #fragment in it, as we do for "any-other-url

  #fragment" if we have cached a document which represents "any-other-url"

  and which is marked no-cache.  Following a href="" link should just result

  in repositioning to the start of the document, not in a new retrieval for

  no-cache documents.

  If the current document is the result of a POST submission, there is in

  general no way to refer to its underlying resource with a URL, since there

  is no URL notation for "url+post_data".  Internal links can be used within

  the context of this document, but there is no way to hyperlink to it or

  parts of it from "outside".  A link to a href="url" or href="url#fragment"

  target should always imply a GET retrieval, whether it occurs within a

  document from a POST submission or not, for consistency.

  This understanding also applies to other than  uses of URL-

  References in documents, for example the link implied between a USEMAP

  attribute and a MAP element; but not to all uses, for example a FORM's

  ACTION requires an external resource and not an internal reference, so

  ACTION="" is resolved to an absolute URL (using a base URL if one is

  given).

  The semantics for URL-less URL-References are given in Internet Draft

  <http://www.ics.uci.edu/~fielding/url/draft-fielding-url-syntax-09.txt>,

  including that "Traversal of such a reference should not result in an

  additional retrieval action."  No comparable prescriptions are made for

  references which have a URL, in particular it is not stated that they

  should be treated as equivalent when the URL coincides with or resolves

  to the current document's retrieval URL.

  The changes to the Lynx code attempt to implement what is outlined above,

  by keeping track of how a link destination was originally specified. href=

  "#fragment" and href="" links are still "resolved" into a "url#fragment"

  and "url" form, for use internally, and if an explicit NOCACHE action is

  requested, and for generating lists of links etc., and in general for use

  outside of the context of the current document.

  To the best of my knowledge they do not lead to unwanted (or misdirected)

  resubmission of POST content without prompting the user for confirmation

  (although the prompt texts probably need improvement). - KW

* The notion of the context of the "current document" is extended, so that

  auxiliary LYNXIMGMAP documents generated by Lynx are within the scope of

  the underlying document, i.e., the document containing the MAP and AREA

  elements.  Although they are treated as separate documents for display

  and internal handling, following links between the underlying document

  and its LYNXIMGMAP documents will result in display from cached data

  (if available) even for no-cache documents, if a fragment-only (or empty)

  URL-Reference was used in the HTML.  This includes the link generated for

  pointing from the IMG to the LYNXIMGMAP if the USEMAP attribute value was

  was fragment-only; the LYNXIMGMAP is normally regenerated from the list

  of MAP and AREA elements kept internally when a link to it is activated.

  (Note that a LYNXIMGMAP screen is still not reloaded from external sources

  even when RELOAD or NOCACHE keys are used, if data for the MAP is found

  in the internal list; but that reloading the underlying document will

  refresh the internal lists.) - KW

* List Page screens are also in the context of the underlying document, so

  that internal links in the underlying document can be listed and activated

  from the List Page.  When following an internal link from a List Page, it

  is not pushed on the history stack (similar to the History List Page), to

  reduce the chance of confusion when walking back through history. - KW

* LYNXIMGMAP and List Page auxiliary documents can now be associated with

  POST data.  This does not mean that the data is posted to a "LYNXIMGMAP:"

  or "file:" URL, which is impossible; but that the underlying document is

  associated with the POST data.  This way these aux documents can still

  have links regarded as "internal" pointing to the document of which they

  are logically a part, keeping track of what POST data are applicable if

  several form submissions have occurred.  Normally the underlying document

  should be in the document cache when such a link is activated, so that

  resubmission of the POST can be overridden.  If it is not found,

  resubmission may happen after user confirmation.  Although the POST data

  content is not shown on such aux pages, it can be inspected with the INFO

  ('=') command.  Note that the association of LYNXIMGMAPs and List Pages

  with POST data automatically excludes them from various operations which

  may otherwise be possible, such as bookmarking or inclusion in the

  'V'isited Links Page.  Note also that such aux pages are treated internally

  as different documents from their POST-free counterparts with the same

  address, which might be generated for example after typing the address

  at a 'g'oto prompt while viewing a different text.  It may be possible to

  generate confusing situations, but regenerating the displayed data (by

  invoking 'L'IST again, or by following a LYNXIMGMAP: link) will always

  show the currently applicable information.

  For image maps, the change implements link-following capabilities which

  come naturally to graphical UAs, which can handle IMG USEMAP and the

  corresponding MAP elements from the same data stream as objects within

  the same document instance. - KW

* To implement the above, MAP and AREA information from a POST response is

  not kept in the global list shared by all other documents, but in a separate

  list specific to the underlying (containing) document and the POST data.

  The specific lists are managed with the underlying document's anchor

  structure so that they can automatically expire after the anchor is removed

  from memory and will not stay around and accumulate until session end. - KW

* POST data is kept in mainloop when following an internal link, but is now

  always dropped for links that were given with a URL.  This removes the

  limitation by which a link in a POST response to the URL of the POST

  submission script would be inappropriately converted to a resubmission of

  the POST data.  (example: search form from 

  calendar>) - KW

* Changes to the PREV_DOC handling in mainloop.  When encountering documents

  that may need to be reloaded while walking back in history, because they

  have POST data associated and are not marked "safe" and are either not found

  in the memory cache or -resubmit_posts has been set, the user is prompted

  for confirmation as before; if the reply is 'n' the document will be skipped

  but if the user cancels with ^G the walkback operation is cancelled and the

  currently viewed document remains loaded.  Some of the aux documents which

  can new be associated with POST data may be automatically skipped if their

  document structures are not any more in memory. - KW

* When following a link marked internal from a List Page and confirmation is

  required, ^G may also have the different effect of cancelling the operation

  while responding 'n' may attempt to follow the link as a GET request (i.e.

  with POST data dropped). - KW

* In some situations, automatic popping from the history stack in mainloop

  after getfile fails to load a new document could lead to resubmission of

  a previous POST request without prompting.  This is now avoided by jumping

  to the code handling PREV_DOC. - KW

* Use new LYinternal_flag variable for propagating info to HTAccess, rather

  than overloading LYoverride_no_cache. - KW

* Tweaks for bad form tag soup with crashes on unclosed SELECT.  Changes in

  HTML.c and GridText.c to prevent memory leaks resulting from insanely

  placed or unclosed form field tags, hopefully not introducing other

  problems... - KW

1997-11-02

* Changes in HTParse's scan to leave additional '#' characters alone after

  the first has been found when scanning (from left) for fragments, and

  to apply the "don't treat '#' as start of a fragment for some kinds of

  URLs" hack only when the '#' will become part of what has been scanned

  as a path.  NNTP articles with more than '#' in the message-id are now

  accessible with the "news:" or "nntp:" schemes.  We should actually always

  escape when generating URLs and unescape when parsing them instead. - KW

* Changed HTUnEscape to leave '%' alone if not followed by two valid hex

  digits. - KW

* Allow the HEAD key and -head and -mime_header flags for "lynxcgi:" URLs

  and for "news:" and "nntp:" URLs that specify a single news article. - KW

* Change to writing of POST data in LYCgi.c, to close file descriptor after

  done writing and before reading as suggested by mhw@bcs.org.uk (Mark

  H. Wilkinson), and to check status from the write() call and retry if

  needed.  Note that the code is not appropriate for handling large

  amounts of POST data, since it still first does all the writing and

  then all the reading. - KW

1997-11-01

* Added support for presentation of text/plain files in SOURCE ('\') mode,

  homologously to text/html files, rather than treating the command as a

  synonym for 'd'ownload when viewing a text/plain file.  Note that if the

  text/plain file has no long lines, the display of it will not change,

  but otherwise, the long lines will not be wrapped in SOURCE mode, and

  thus can be 'p'rinted without line wrapping.  One still should use

  'd'ownload for a faithful copy of the file, i.e., without charset

  conversions and expansions of TABs. - FM

1997-10-30

* Get numbering of links on List Page right, by faking hidden links for

  form fields which get counted but are not selectable. - KW

* Allow LYNXHIST: links in List Page since the History Page may get

  'L'isted... - KW

* Mods of HTML.c parsing to close any open Anchor or emphasis elements

  at the starts of TABLE elements, as the Big Two's TABLE handlers appear

  to do now. - FM

* Tweaks of charset handling via META elements and CHARSET attributes. - FM

1997-10-29

* Tweak of SortaSGML DTD for insane HTML with FORM start before HEAD. - KW

* Prevent memory leaks from HTChunks for option and textarea with very

  bad markup, added TRACE and BAD_HTML_USE_TRACE warnings for those. - KW

* Don't call tigetnum("ncv") for _WINDOWS - BD

* Added Unicode tables for the "DEC Multinational", "Macintosh (8 bit)",

  and "NeXT character set" display character sets. - FM

1997-10-28

* Disable setting of the `reloading' flag in HTLoadDocument which would

  generate no-cache headers for request where this is not appropriate. - KW

* Memory leak for first OPTION in SELECT removed. - KW

* Changed HTDOS_name to not return pointer to static area. - Binh Do

  <bdo@mailman.epnet.com>

* Changes to INSTALLATION file. - HN

* Fix in HTMIME.c to recognize Content-location. - KW

1997-10-27

* Don't disable FORMs that have INPUTs with TYPE="file" or TYPE="range"

  since those fields may be optional and the rest of the FORM may be

  valid to submit. - FM

* Updated links to www.w3.org in help files. - KW

* Some changes for DOS compilation. - DK

* Changes to INSTALLATION file. - HN

* Tweaks of internal links stuff. - KW

* Allow HEAD on URLs starting with LYNXIMGMAP:http by stripping the

  prefix. - KW

* Reverted LYCheckForCSI() to previous incarnation. - KW

1997-10-26

* If compiled with -DDONT_TRACK_INTERNAL_LINKS, behavior with respect to

  internal links as in FM's code. - KW

* Some chartrans tweaks. - KW

1997-10-25

* Changes to INSTALLATION file. - HN

* Make temp file names conform to 8+3 naming convention for DOS, and

  give LYNX_TEMP_SPACE priority over TEMP and TMP for DOS. - DK

* Create temp file names with '-' char as separator between pid and

  sequential number for non-8+3 systems, to prevent (unlikely)

  collision. - KW

1997-10-24

* Change in HTML.c to use the base (rather than current document, if the

  base differs) in all cases for form ACTIONs (i.e., if the ACTION is

  empty, or has a lone fragment, or an ACTION attribute is not included)