/*
 * $LynxId: LYCharUtils.c,v 1.110 2011/06/04 14:18:10 tom Exp $
 *
 *  Functions associated with LYCharSets.c and the Lynx version of HTML.c - FM
 *  ==========================================================================
 */
#include <HTUtils.h>
#include <SGML.h>

#define Lynx_HTML_Handler
#include <HTChunk.h>
#include <HText.h>
#include <HTStyle.h>
#include <HTMIME.h>
#include <HTML.h>

#include <HTCJK.h>
#include <HTAtom.h>
#include <HTMLGen.h>
#include <HTParse.h>
#include <UCMap.h>
#include <UCDefs.h>
#include <UCAux.h>

#include <LYGlobalDefs.h>
#include <LYCharUtils.h>
#include <LYCharSets.h>

#include <HTAlert.h>
#include <HTForms.h>
#include <HTNestedList.h>
#include <GridText.h>
#include <LYStrings.h>
#include <LYUtils.h>
#include <LYMap.h>
#include <LYBookmark.h>
#include <LYCurses.h>
#include <LYCookie.h>

#include <LYexit.h>
#include <LYLeaks.h>

/*
 * Used for nested lists.  - FM
 */
int OL_CONTINUE = -29999;	/* flag for whether CONTINUE is set */
int OL_VOID = -29998;		/* flag for whether a count is set */

/*
 *  This function converts any ampersands in allocated
 *  strings to "&amp;".  If isTITLE is TRUE, it also
 *  converts any angle-brackets to "&lt;" or "&gt;". - FM
 */
void LYEntify(char **str,
	      int isTITLE)
{
    char *p = *str;
    char *q = NULL, *cp = NULL;
    int amps = 0, lts = 0, gts = 0;

#ifdef CJK_EX
    enum _state {
	S_text,
	S_esc,
	S_dollar,
	S_paren,
	S_nonascii_text,
	S_dollar_paren
    } state = S_text;
    int in_sjis = 0;
#endif

    if (isEmpty(p))
	return;

    /*
     * Count the ampersands.  - FM
     */
    while ((*p != '\0') && (q = strchr(p, '&')) != NULL) {
	amps++;
	p = (q + 1);
    }

    /*
     * Count the left-angle-brackets, if needed.  - FM
     */
    if (isTITLE == TRUE) {
	p = *str;
	while ((*p != '\0') && (q = strchr(p, '<')) != NULL) {
	    lts++;
	    p = (q + 1);
	}
    }

    /*
     * Count the right-angle-brackets, if needed.  - FM
     */
    if (isTITLE == TRUE) {
	p = *str;
	while ((*p != '\0') && (q = strchr(p, '>')) != NULL) {
	    gts++;
	    p = (q + 1);
	}
    }

    /*
     * Check whether we need to convert anything.  - FM
     */
    if (amps == 0 && lts == 0 && gts == 0)
	return;

    /*
     * Allocate space and convert.  - FM
     */
    q = typecallocn(char,
		    (strlen(*str)
		     + (unsigned)(4 * amps)
		     + (unsigned)(3 * lts)
		     + (unsigned)(3 * gts) + 1));
    if ((cp = q) == NULL)
	outofmem(__FILE__, "LYEntify");

    assert(cp != NULL);
    assert(q != NULL);

    for (p = *str; *p; p++) {
#ifdef CJK_EX
	if (IS_CJK_TTY) {
	    switch (state) {
	    case S_text:
		if (*p == '\033') {
		    state = S_esc;
		    *q++ = *p;
		    continue;
		}
		break;

	    case S_esc:
		if (*p == '$') {
		    state = S_dollar;
		    *q++ = *p;
		    continue;
		} else if (*p == '(') {
		    state = S_paren;
		    *q++ = *p;
		    continue;
		} else {
		    state = S_text;
		    *q++ = *p;
		    continue;
		}

	    case S_dollar:
		if (*p == '@' || *p == 'B' || *p == 'A') {
		    state = S_nonascii_text;
		    *q++ = *p;
		    continue;
		} else if (*p == '(') {
		    state = S_dollar_paren;
		    *q++ = *p;
		    continue;
		} else {
		    state = S_text;
		    *q++ = *p;
		    continue;
		}

	    case S_dollar_paren:
		if (*p == 'C') {
		    state = S_nonascii_text;
		    *q++ = *p;
		    continue;
		} else {
		    state = S_text;
		    *q++ = *p;
		    continue;
		}

	    case S_paren:
		if (*p == 'B' || *p == 'J' || *p == 'T') {
		    state = S_text;
		    *q++ = *p;
		    continue;
		} else if (*p == 'I') {
		    state = S_nonascii_text;
		    *q++ = *p;
		    continue;
		}
		/* FALLTHRU */

	    case S_nonascii_text:
		if (*p == '\033')
		    state = S_esc;
		*q++ = *p;
		continue;

	    default:
		break;
	    }
	    if (*(p + 1) != '\0' &&
		(IS_EUC(UCH(*p), UCH(*(p + 1))) ||
		 IS_SJIS(UCH(*p), UCH(*(p + 1)), in_sjis) ||
		 IS_BIG5(UCH(*p), UCH(*(p + 1))))) {
		*q++ = *p++;
		*q++ = *p;
		continue;
	    }
	}
#endif
	if (*p == '&') {
	    *q++ = '&';
	    *q++ = 'a';
	    *q++ = 'm';
	    *q++ = 'p';
	    *q++ = ';';
	} else if (isTITLE && *p == '<') {
	    *q++ = '&';
	    *q++ = 'l';
	    *q++ = 't';
	    *q++ = ';';
	} else if (isTITLE && *p == '>') {
	    *q++ = '&';
	    *q++ = 'g';
	    *q++ = 't';
	    *q++ = ';';
	} else {
	    *q++ = *p;
	}
    }
    *q = '\0';
    FREE(*str);
    *str = cp;
}

/*
 * Callers to LYEntifyTitle/LYEntifyValue do not look at the 'target' param.
 * Optimize things a little by avoiding the memory allocation if not needed,
 * as is usually the case.
 */
static BOOL MustEntify(const char *source)
{
    BOOL result;

#ifdef CJK_EX
    if (IS_CJK_TTY && strchr(source, '\033') != 0) {
	result = TRUE;
    } else
#endif
    {
	size_t length = strlen(source);
	size_t reject = strcspn(source, "<&>");

	result = (BOOL) (length != reject);
    }

    return result;
}

/*
 * Wrappers for LYEntify() which do not assume that the source was allocated,
 * e.g., output from gettext().
 */
const char *LYEntifyTitle(char **target, const char *source)
{
    const char *result = 0;

    if (MustEntify(source)) {
	StrAllocCopy(*target, source);
	LYEntify(target, TRUE);
	result = *target;
    } else {
	result = source;
    }
    return result;
}

const char *LYEntifyValue(char **target, const char *source)
{
    const char *result = 0;

    if (MustEntify(source)) {
	StrAllocCopy(*target, source);
	LYEntify(target, FALSE);
	result = *target;
    } else {
	result = source;
    }
    return result;
}

/*
 *  This function trims characters <= that of a space (32),
 *  including HT_NON_BREAK_SPACE (1) and HT_EN_SPACE (2),
 *  but not ESC, from the heads of strings. - FM
 */
void LYTrimHead(char *str)
{
    const char *s = str;

    if (isEmpty(s))
	return;

    while (*s && WHITE(*s) && UCH(*s) != UCH(CH_ESC))	/* S/390 -- gil -- 1669 */
	s++;
    if (s > str) {
	char *ns = str;

	while (*s) {
	    *ns++ = *s++;
	}
	*ns = '\0';
    }
}

/*
 *  This function trims characters <= that of a space (32),
 *  including HT_NON_BREAK_SPACE (1), HT_EN_SPACE (2), and
 *  ESC from the tails of strings. - FM
 */
void LYTrimTail(char *str)
{
    int i;

    if (isEmpty(str))
	return;

    i = (int) strlen(str) - 1;
    while (i >= 0) {
	if (WHITE(str[i]))
	    str[i] = '\0';
	else
	    break;
	i--;
    }
}

/*
 * This function should receive a pointer to the start
 * of a comment.  It returns a pointer to the end ('>')
 * character of comment, or it's best guess if the comment
 * is invalid. - FM
 */
char *LYFindEndOfComment(char *str)
{
    char *cp, *cp1;
    enum comment_state {
	start1,
	start2,
	end1,
	end2
    } state;

    if (str == NULL)
	/*
	 * We got NULL, so return NULL.  - FM
	 */
	return NULL;

    if (StrNCmp(str, "<!--", 4))
	/*
	 * We don't have the start of a comment, so return the beginning of the
	 * string.  - FM
	 */
	return str;

    cp = (str + 4);
    if (*cp == '>')
	/*
	 * It's an invalid comment, so
	 * return this end character. - FM
	 */
	return cp;

    if ((cp1 = strchr(cp, '>')) == NULL)
	/*
	 * We don't have an end character, so return the beginning of the
	 * string.  - FM
	 */
	return str;

    if (*cp == '-')
	/*
	 * Ugh, it's a "decorative" series of dashes, so return the next end
	 * character.  - FM
	 */
	return cp1;

    /*
     * OK, we're ready to start parsing.  - FM
     */
    state = start2;
    while (*cp != '\0') {
	switch (state) {
	case start1:
	    if (*cp == '-')
		state = start2;
	    else
		/*
		 * Invalid comment, so return the first '>' from the start of
		 * the string.  - FM
		 */
		return cp1;
	    break;

	case start2:
	    if (*cp == '-')
		state = end1;
	    break;

	case end1:
	    if (*cp == '-')
		state = end2;
	    else
		/*
		 * Invalid comment, so return the first '>' from the start of
		 * the string.  - FM
		 */
		return cp1;
	    break;

	case end2:
	    if (*cp == '>')
		/*
		 * Valid comment, so return the end character.  - FM
		 */
		return cp;
	    if (*cp == '-') {
		state = start1;
	    } else if (!(WHITE(*cp) && UCH(*cp) != UCH(CH_ESC))) {	/* S/390 -- gil -- 1686 */
		/*
		 * Invalid comment, so return the first '>' from the start of
		 * the string.  - FM
		 */
		return cp1;
	    }
	    break;

	default:
	    break;
	}
	cp++;
    }

    /*
     * Invalid comment, so return the first '>' from the start of the string. 
     * - FM
     */
    return cp1;
}

/*
 *  If an HREF, itself or if resolved against a base,
 *  represents a file URL, and the host is defaulted,
 *  force in "//localhost".  We need this until
 *  all the other Lynx code which performs security
 *  checks based on the "localhost" string is changed
 *  to assume "//localhost" when a host field is not
 *  present in file URLs - FM
 */
void LYFillLocalFileURL(char **href,
			const char *base)
{
    char *temp = NULL;

    if (isEmpty(*href))
	return;

    if (!strcmp(*href, "//") || !StrNCmp(*href, "///", 3)) {
	if (base != NULL && isFILE_URL(base)) {
	    StrAllocCopy(temp, STR_FILE_URL);
	    StrAllocCat(temp, *href);
	    StrAllocCopy(*href, temp);
	}
    }
    if (isFILE_URL(*href)) {
	if (*(*href + 5) == '\0') {
	    StrAllocCat(*href, "//localhost");
	} else if (!strcmp(*href, "file://")) {
	    StrAllocCat(*href, "localhost");
	} else if (!StrNCmp(*href, "file:///", 8)) {
	    StrAllocCopy(temp, (*href + 7));
	    LYLocalFileToURL(href, temp);
	} else if (!StrNCmp(*href, "file:/", 6) && !LYIsHtmlSep(*(*href + 6))) {
	    StrAllocCopy(temp, (*href + 5));
	    LYLocalFileToURL(href, temp);
	}
    }
#if defined(USE_DOS_DRIVES)
    if (LYIsDosDrive(*href)) {
	/*
	 * If it's a local DOS path beginning with drive letter,
	 * add file://localhost/ prefix and go ahead.
	 */
	StrAllocCopy(temp, *href);
	LYLocalFileToURL(href, temp);
    }

    /* use below: strlen("file://localhost/") = 17 */
    if (!StrNCmp(*href, "file://localhost/", 17)
	&& (strlen(*href) == 19)
	&& LYIsDosDrive(*href + 17)) {
	/*
	 * Terminate DOS drive letter with a slash to surf root successfully.
	 * Here seems a proper place to do so.
	 */
	LYAddPathSep(href);
    }
#endif /* USE_DOS_DRIVES */

    /*
     * No path in a file://localhost URL means a
     * directory listing for the current default. - FM
     */
    if (!strcmp(*href, "file://localhost")) {
	const char *temp2;

#ifdef VMS
	temp2 = HTVMS_wwwName(LYGetEnv("PATH"));
#else
	char curdir[LY_MAXPATH];

	temp2 = wwwName(Current_Dir(curdir));
#endif /* VMS */
	if (!LYIsHtmlSep(*temp2))
	    LYAddHtmlSep(href);
	/*
	 * Check for pathological cases - current dir has chars which MUST BE
	 * URL-escaped - kw
	 */
	if (strchr(temp2, '%') != NULL || strchr(temp2, '#') != NULL) {
	    FREE(temp);
	    temp = HTEscape(temp2, URL_PATH);
	    StrAllocCat(*href, temp);
	} else {
	    StrAllocCat(*href, temp2);
	}
    }
#ifdef VMS
    /*
     * On VMS, a file://localhost/ URL means
     * a listing for the login directory. - FM
     */
    if (!strcmp(*href, "file://localhost/"))
	StrAllocCat(*href, (HTVMS_wwwName(Home_Dir()) + 1));
#endif /* VMS */

    FREE(temp);
    return;
}

/*
 *  This function writes a line with a META tag to an open file,
 *  which will specify a charset parameter to use when the file is
 *  read back in.  It is meant for temporary HTML files used by the
 *  various special pages which may show titles of documents.  When those
 *  files are created, the title strings normally have been translated and
 *  expanded to the display character set, so we have to make sure they
 *  don't get translated again.
 *  If the user has changed the display character set during the lifetime
 *  of the Lynx session (or, more exactly, during the time the title
 *  strings to be written were generated), they may now have different
 *  character encodings and there is currently no way to get it all right.
 *  To change this, we would have to add a variable for each string which
 *  keeps track of its character encoding.
 *  But at least we can try to ensure that reading the file after future
 *  display character set changes will give reasonable output.
 *
 *  The META tag is not written if the display character set (passed as
 *  disp_chndl) already corresponds to the charset assumption that
 *  would be made when the file is read. - KW
 *
 *  Currently this function is used for temporary files like "Lynx Info Page"
 *  and for one permanent - bookmarks (so it may be a problem if you change
 *  the display charset later: new bookmark entries may be mistranslated).
 *								 - LP
 */
void LYAddMETAcharsetToFD(FILE *fd, int disp_chndl)
{
    if (disp_chndl == -1)
	/*
	 * -1 means use current_char_set.
	 */
	disp_chndl = current_char_set;

    if (fd == NULL || disp_chndl < 0)
	/*
	 * Should not happen.
	 */
	return;

    if (UCLYhndl_HTFile_for_unspec == disp_chndl)
	/*
	 * Not need to do, so we don't.
	 */
	return;

    if (LYCharSet_UC[disp_chndl].enc == UCT_ENC_7BIT)
	/*
	 * There shouldn't be any 8-bit characters in this case.
	 */
	return;

    /*
     * In other cases we don't know because UCLYhndl_for_unspec may change
     * during the lifetime of the file (by toggling raw mode or changing the
     * display character set), so proceed.
     */
    fprintf(fd, "<META %s content=\"text/html;charset=%s\">\n",
	    "http-equiv=\"content-type\"",
	    LYCharSet_UC[disp_chndl].MIMEname);
}

/*
 * This function returns OL TYPE="A" strings in
 * the range of " A." (1) to "ZZZ." (18278). - FM
 */
char *LYUppercaseA_OL_String(int seqnum)
{
    static char OLstring[8];

    if (seqnum <= 1) {
	strcpy(OLstring, " A.");
	return OLstring;
    }
    if (seqnum < 27) {
	sprintf(OLstring, " %c.", (seqnum + 64));
	return OLstring;
    }
    if (seqnum < 703) {
	sprintf(OLstring, "%c%c.", ((seqnum - 1) / 26 + 64),
		(seqnum - ((seqnum - 1) / 26) * 26 + 64));
	return OLstring;
    }
    if (seqnum < 18279) {
	sprintf(OLstring, "%c%c%c.", ((seqnum - 27) / 676 + 64),
		(((seqnum - ((seqnum - 27) / 676) * 676) - 1) / 26 + 64),
		(seqnum - ((seqnum - 1) / 26) * 26 + 64));
	return OLstring;
    }
    strcpy(OLstring, "ZZZ.");
    return OLstring;
}

/*
 * This function returns OL TYPE="a" strings in
 * the range of " a." (1) to "zzz." (18278). - FM
 */
char *LYLowercaseA_OL_String(int seqnum)
{
    static char OLstring[8];

    if (seqnum <= 1) {
	strcpy(OLstring, " a.");
	return OLstring;
    }
    if (seqnum < 27) {
	sprintf(OLstring, " %c.", (seqnum + 96));
	return OLstring;
    }
    if (seqnum < 703) {
	sprintf(OLstring, "%c%c.", ((seqnum - 1) / 26 + 96),
		(seqnum - ((seqnum - 1) / 26) * 26 + 96));
	return OLstring;
    }
    if (seqnum < 18279) {
	sprintf(OLstring, "%c%c%c.", ((seqnum - 27) / 676 + 96),
		(((seqnum - ((seqnum - 27) / 676) * 676) - 1) / 26 + 96),
		(seqnum - ((seqnum - 1) / 26) * 26 + 96));
	return OLstring;
    }
    strcpy(OLstring, "zzz.");
    return OLstring;
}

/*
 * This function returns OL TYPE="I" strings in the
 * range of " I." (1) to "MMM." (3000).- FM
 * Maximum length: 16 -TD
 */
char *LYUppercaseI_OL_String(int seqnum)
{
    static char OLstring[20];
    int Arabic = seqnum;

    if (Arabic >= 3000) {
	strcpy(OLstring, "MMM.");
	return OLstring;
    }

    switch (Arabic) {
    case 1:
	strcpy(OLstring, " I.");
	return OLstring;
    case 5:
	strcpy(OLstring, " V.");
	return OLstring;
    case 10:
	strcpy(OLstring, " X.");
	return OLstring;
    case 50:
	strcpy(OLstring, " L.");
	return OLstring;
    case 100:
	strcpy(OLstring, " C.");
	return OLstring;
    case 500:
	strcpy(OLstring, " D.");
	return OLstring;
    case 1000:
	strcpy(OLstring, " M.");
	return OLstring;
    default:
	OLstring[0] = '\0';
	break;
    }

    while (Arabic >= 1000) {
	strcat(OLstring, "M");
	Arabic -= 1000;
    }

    if (Arabic >= 900) {
	strcat(OLstring, "CM");
	Arabic -= 900;
    }

    if (Arabic >= 500) {
	strcat(OLstring, "D");
	Arabic -= 500;
	while (Arabic >= 500) {
	    strcat(OLstring, "C");
	    Arabic -= 10;
	}
    }

    if (Arabic >= 400) {
	strcat(OLstring, "CD");
	Arabic -= 400;
    }

    while (Arabic >= 100) {
	strcat(OLstring, "C");
	Arabic -= 100;
    }

    if (Arabic >= 90) {
	strcat(OLstring, "XC");
	Arabic -= 90;
    }

    if (Arabic >= 50) {
	strcat(OLstring, "L");
	Arabic -= 50;
	while (Arabic >= 50) {
	    strcat(OLstring, "X");
	    Arabic -= 10;
	}
    }

    if (Arabic >= 40) {
	strcat(OLstring, "XL");
	Arabic -= 40;
    }

    while (Arabic > 10) {
	strcat(OLstring, "X");
	Arabic -= 10;
    }

    switch (Arabic) {
    case 1:
	strcat(OLstring, "I.");
	break;
    case 2:
	strcat(OLstring, "II.");
	break;
    case 3:
	strcat(OLstring, "III.");
	break;
    case 4:
	strcat(OLstring, "IV.");
	break;
    case 5:
	strcat(OLstring, "V.");
	break;
    case 6:
	strcat(OLstring, "VI.");
	break;
    case 7:
	strcat(OLstring, "VII.");
	break;
    case 8:
	strcat(OLstring, "VIII.");
	break;
    case 9:
	strcat(OLstring, "IX.");
	break;
    case 10:
	strcat(OLstring, "X.");
	break;
    default:
	strcat(OLstring, ".");
	break;
    }

    return OLstring;
}

/*
 * This function returns OL TYPE="i" strings in
 * range of " i." (1) to "mmm." (3000).- FM
 * Maximum length: 16 -TD
 */
char *LYLowercaseI_OL_String(int seqnum)
{
    static char OLstring[20];
    int Arabic = seqnum;

    if (Arabic >= 3000) {
	strcpy(OLstring, "mmm.");
	return OLstring;
    }

    switch (Arabic) {
    case 1:
	strcpy(OLstring, " i.");
	return OLstring;
    case 5:
	strcpy(OLstring, " v.");
	return OLstring;
    case 10:
	strcpy(OLstring, " x.");
	return OLstring;
    case 50:
	strcpy(OLstring, " l.");
	return OLstring;
    case 100:
	strcpy(OLstring, " c.");
	return OLstring;
    case 500:
	strcpy(OLstring, " d.");
	return OLstring;
    case 1000:
	strcpy(OLstring, " m.");
	return OLstring;
    default:
	OLstring[0] = '\0';
	break;
    }

    while (Arabic >= 1000) {
	strcat(OLstring, "m");
	Arabic -= 1000;
    }

    if (Arabic >= 900) {
	strcat(OLstring, "cm");
	Arabic -= 900;
    }

    if (Arabic >= 500) {
	strcat(OLstring, "d");
	Arabic -= 500;
	while (Arabic >= 500) {
	    strcat(OLstring, "c");
	    Arabic -= 10;
	}
    }

    if (Arabic >= 400) {
	strcat(OLstring, "cd");
	Arabic -= 400;
    }

    while (Arabic >= 100) {
	strcat(OLstring, "c");
	Arabic -= 100;
    }

    if (Arabic >= 90) {
	strcat(OLstring, "xc");
	Arabic -= 90;
    }

    if (Arabic >= 50) {
	strcat(OLstring, "l");
	Arabic -= 50;
	while (Arabic >= 50) {
	    strcat(OLstring, "x");
	    Arabic -= 10;
	}
    }

    if (Arabic >= 40) {
	strcat(OLstring, "xl");
	Arabic -= 40;
    }

    while (Arabic > 10) {
	strcat(OLstring, "x");
	Arabic -= 10;
    }

    switch (Arabic) {
    case 1:
	strcat(OLstring, "i.");
	break;
    case 2:
	strcat(OLstring, "ii.");
	break;
    case 3:
	strcat(OLstring, "iii.");
	break;
    case 4:
	strcat(OLstring, "iv.");
	break;
    case 5:
	strcat(OLstring, "v.");
	break;
    case 6:
	strcat(OLstring, "vi.");
	break;
    case 7:
	strcat(OLstring, "vii.");
	break;
    case 8:
	strcat(OLstring, "viii.");
	break;
    case 9:
	strcat(OLstring, "ix.");
	break;
    case 10:
	strcat(OLstring, "x.");
	break;
    default:
	strcat(OLstring, ".");
	break;
    }

    return OLstring;
}

/*
 *  This function initializes the Ordered List counter. - FM
 */
void LYZero_OL_Counter(HTStructured * me)
{
    int i;

    if (!me)
	return;

    for (i = 0; i < 12; i++) {
	me->OL_Counter[i] = OL_VOID;
	me->OL_Type[i] = '1';
    }

    me->Last_OL_Count = 0;
    me->Last_OL_Type = '1';

    return;
}

/*
 *  This function is used by the HTML Structured object. - KW
 */
void LYGetChartransInfo(HTStructured * me)
{
    me->UCLYhndl = HTAnchor_getUCLYhndl(me->node_anchor,
					UCT_STAGE_STRUCTURED);
    if (me->UCLYhndl < 0) {
	int chndl = HTAnchor_getUCLYhndl(me->node_anchor, UCT_STAGE_HTEXT);

	if (chndl < 0) {
	    chndl = current_char_set;
	    HTAnchor_setUCInfoStage(me->node_anchor, chndl,
				    UCT_STAGE_HTEXT,
				    UCT_SETBY_STRUCTURED);
	}
	HTAnchor_setUCInfoStage(me->node_anchor, chndl,
				UCT_STAGE_STRUCTURED,
				UCT_SETBY_STRUCTURED);
	me->UCLYhndl = HTAnchor_getUCLYhndl(me->node_anchor,
					    UCT_STAGE_STRUCTURED);
    }
    me->UCI = HTAnchor_getUCInfoStage(me->node_anchor,
				      UCT_STAGE_STRUCTURED);
}

	/* as in HTParse.c, saves some calls - kw */
static const char *hex = "0123456789ABCDEF";

/*
 *	  Any raw 8-bit or multibyte characters already have been
 *	  handled in relation to the display character set
 *	  in SGML_character(), including named and numeric entities.
 *
 *  This function used for translations HTML special fields inside tags
 *  (ALT=, VALUE=, etc.) from charset `cs_from' to charset `cs_to'.
 *  It also unescapes non-ASCII characters from URL (#fragments !)
 *  if st_URL is active.
 *
 *  If `do_ent' is YES, it converts named entities
 *  and numeric character references (NCRs) to their `cs_to' replacements.
 *
 *  Named entities converted to unicodes.  NCRs (unicodes) converted
 *  by UCdomap.c chartrans functions.
 *  ???NCRs with values in the ISO-8859-1 range 160-255 may be converted
 *  to their HTML entity names (via old-style entities) and then translated
 *  according to the LYCharSets.c array for `cs_out'???.
 *
 *  Some characters (see descriptions in `put_special_unicodes' from SGML.c)
 *  translated in relation with the state of boolean variables
 *  `use_lynx_specials', `plain_space' and `hidden'.  It is not clear yet:
 *
 *  If plain_space is TRUE, nbsp (160) will be treated as an ASCII
 *  space (32).  If hidden is TRUE, entities will be translated
 *  (if `do_ent' is YES) but escape sequences will be passed unaltered.
 *  If `hidden' is FALSE, some characters are converted to Lynx special
 *  codes (see `put_special_unicodes') or ASCII space if `plain_space'
 *  applies).  @@ is `use_lynx_specials' needed, does it have any effect? @@
 *  If `use_lynx_specials' is YES, translate byte values 160 and 173
 *  meaning U+00A0 and U+00AD given as or converted from raw char input
 *  are converted to HT_NON_BREAK_SPACE and LY_SOFT_HYPHEN, respectively
 *  (unless input and output charset are both iso-8859-1, for compatibility
 *  with previous usage in HTML.c) even if `hidden' or `plain_space' is set.
 *
 *  If `Back' is YES, the reverse is done instead i.e., Lynx special codes
 *  in the input are translated back to character values.
 *
 *  If `Back' is YES, an attempt is made to use UCReverseTransChar() for
 *  back translation which may be more efficient. (?)
 *
 *  If `stype' is st_URL, non-ASCII characters are URL-encoded instead.
 *  The sequence of bytes being URL-encoded is the raw input character if
 *  we couldn't translate it from `cs_in' (CJK etc.); otherwise it is the
 *  UTF-8 representation if either `cs_to' requires this or if the
 *  character's Unicode value is > 255, otherwise it should be the iso-8859-1
 *  representation.
 *  No general URL-encoding occurs for displayable ASCII characters and
 *  spaces and some C0 controls valid in HTML (LF, TAB), it is expected
 *  that other functions will take care of that as appropriate.
 *
 *  Escape characters (0x1B, '\033') are
 *  - URL-encoded	if `stype'  is st_URL,	 otherwise
 *  - dropped		if `stype'  is st_other, otherwise (i.e., st_HTML)
 *  - passed		if `hidden' is TRUE or HTCJK is set, otherwise
 *  - dropped.
 *
 *  (If `stype' is st_URL or st_other most of the parameters really predefined:
 *  cs_from=cs_to, use_lynx_specials=plain_space=NO, and hidden=YES)
 *
 *
 *  Returns pointer to the char** passed in
 *		 if string translated or translation unnecessary,
 *	    NULL otherwise
 *		 (in which case something probably went wrong.)
 *
 *
 *  In general, this somehow ugly function (KW)
 *  cover three functions from v.2.7.2 (FM):
 *		    extern void LYExpandString (
 *		       HTStructured *	       me,
 *		       char **		       str);
 *		    extern void LYUnEscapeEntities (
 *		       HTStructured *	       me,
 *		       char **		       str);
 *		    extern void LYUnEscapeToLatinOne (
 *		       HTStructured *	       me,
 *		       char **		       str,
 *		       BOOLEAN		       isURL);
 */

char **LYUCFullyTranslateString(char **str,
				int cs_from,
				int cs_to,
				int do_ent,
				int use_lynx_specials,
				int plain_space,
				int hidden,
				int Back,
				CharUtil_st stype)
{
    char *p;
    char *q, *qs;
    HTChunk *chunk = NULL;
    char *cp = 0;
    char cpe = 0;
    char *esc = NULL;
    char replace_buf[64];
    int uck;
    int lowest_8;
    UCode_t code = 0;
    unsigned long lcode;
    BOOL output_utf8 = 0, repl_translated_C0 = 0;
    size_t len;
    const char *name = NULL;
    BOOLEAN no_bytetrans;
    UCTransParams T;
    BOOL from_is_utf8 = FALSE;
    char *puni;
    enum _state {
	S_text,
	S_esc,
	S_dollar,
	S_paren,
	S_nonascii_text,
	S_dollar_paren,
	S_trans_byte,
	S_check_ent,
	S_ncr,
	S_check_uni,
	S_named,
	S_check_name,
	S_recover,
	S_got_oututf8,
	S_got_outstring,
	S_put_urlstring,
	S_got_outchar,
	S_put_urlchar,
	S_next_char,
	S_done
    } state = S_text;
    enum _parsing_what {
	P_text,
	P_utf8,
	P_hex,
	P_decimal,
	P_named
    } what = P_text;

#ifdef KANJI_CODE_OVERRIDE
    static unsigned char sjis_1st = '\0';

    unsigned char sjis_str[3];
#endif

    /*
     * Make sure we have a non-empty string.  - FM
     */
    if (isEmpty(*str))
	return str;

    /*
     * FIXME: something's wrong with the limit checks here (clearing the
     * buffer helps).
     */
    memset(replace_buf, 0, sizeof(replace_buf));

    /*
     * Don't do byte translation if original AND target character sets are both
     * iso-8859-1 (and we are not called to back-translate), or if we are in
     * CJK mode.
     */
    if (IS_CJK_TTY
#ifdef EXP_JAPANESEUTF8_SUPPORT
	&& (strcmp(LYCharSet_UC[cs_from].MIMEname, "utf-8") != 0)
	&& (strcmp(LYCharSet_UC[cs_to].MIMEname, "utf-8") != 0)
#endif
	) {
	no_bytetrans = TRUE;
    } else if (cs_to <= 0 && cs_from == cs_to && (!Back || cs_to < 0)) {
	no_bytetrans = TRUE;
    } else {
	/* No need to translate or examine the string any further */
	no_bytetrans = (BOOL) (!use_lynx_specials && !Back &&
			       UCNeedNotTranslate(cs_from, cs_to));
    }
    /*
     * Save malloc/calloc overhead in simple case - kw
     */
    if (do_ent && hidden && (stype != st_URL) && (strchr(*str, '&') == NULL))
	do_ent = FALSE;

    /* Can't do, caller should figure out what to do... */
    if (!UCCanTranslateFromTo(cs_from, cs_to)) {
	if (cs_to < 0)
	    return NULL;
	if (!do_ent && no_bytetrans)
	    return NULL;
	no_bytetrans = TRUE;
    } else if (cs_to < 0) {
	do_ent = FALSE;
    }

    if (!do_ent && no_bytetrans)
	return str;
    p = *str;

    if (!no_bytetrans) {
	UCTransParams_clear(&T);
	UCSetTransParams(&T, cs_from, &LYCharSet_UC[cs_from],
			 cs_to, &LYCharSet_UC[cs_to]);
	from_is_utf8 = (BOOL) (LYCharSet_UC[cs_from].enc == UCT_ENC_UTF8);
	output_utf8 = T.output_utf8;
	repl_translated_C0 = T.repl_translated_C0;
	puni = p;
    } else if (do_ent) {
	output_utf8 = (BOOL) (LYCharSet_UC[cs_to].enc == UCT_ENC_UTF8 ||
			      HText_hasUTF8OutputSet(HTMainText));
	repl_translated_C0 = (BOOL) (LYCharSet_UC[cs_to].enc == UCT_ENC_8BIT_C0);
    }

    lowest_8 = LYlowest_eightbit[cs_to];

    /*
     * Create a buffer string seven times the length of the original, so we
     * have plenty of room for expansions.  - FM
     */
    len = strlen(p) + 16;
    q = p;

    qs = q;

/*  Create the HTChunk only if we need it */
#define CHUNK (chunk ? chunk : (chunk = HTChunkCreate2(128, len+1)))

#define REPLACE_STRING(s) \
		if (q != qs) HTChunkPutb(CHUNK, qs, (int) (q - qs)); \
		HTChunkPuts(CHUNK, s); \
		qs = q = *str

#define REPLACE_CHAR(c) if (q > p) { \
		HTChunkPutb(CHUNK, qs, (int) (q - qs)); \
		qs = q = *str; \
		*q++ = c; \
	    } else \
		*q++ = c

    /*
     * Loop through string, making conversions as needed.
     *
     * The while() checks for a non-'\0' char only for the normal text states
     * since other states may temporarily modify p or *p (which should be
     * restored before S_done!) - kw
     */
    while (*p || (state != S_text && state != S_nonascii_text)) {
	switch (state) {
	case S_text:
	    code = UCH(*p);
#ifdef KANJI_CODE_OVERRIDE
	    if (HTCJK == JAPANESE && last_kcode == SJIS) {
		if (sjis_1st == '\0' && (IS_SJIS_HI1(code) || IS_SJIS_HI2(code))) {
		    sjis_1st = UCH(code);
		} else if (sjis_1st && IS_SJIS_LO(code)) {
		    sjis_1st = '\0';
		} else {
		    if (conv_jisx0201kana && 0xA1 <= code && code <= 0xDF) {
			sjis_str[2] = '\0';
			JISx0201TO0208_SJIS(UCH(code),
					    sjis_str, sjis_str + 1);
			REPLACE_STRING(sjis_str);
			p++;
			continue;
		    }
		}
	    }
#endif
	    if (*p == '\033') {
		if ((IS_CJK_TTY && !hidden) || stype != st_HTML) {
		    state = S_esc;
		    if (stype == st_URL) {
			REPLACE_STRING("%1B");
			p++;
			continue;
		    } else if (stype != st_HTML) {
			p++;
			continue;
		    } else {
			*q++ = *p++;
			continue;
		    }
		} else if (!hidden) {
		    /*
		     * CJK handling not on, and not a hidden INPUT, so block
		     * escape.  - FM
		     */
		    state = S_next_char;
		} else {
		    state = S_trans_byte;
		}
	    } else {
		state = (do_ent ? S_check_ent : S_trans_byte);
	    }
	    break;

	case S_esc:
	    if (*p == '$') {
		state = S_dollar;
		*q++ = *p++;
		continue;
	    } else if (*p == '(') {
		state = S_paren;
		*q++ = *p++;
		continue;
	    } else {
		state = S_text;
	    }
	    break;

	case S_dollar:
	    if (*p == '@' || *p == 'B' || *p == 'A') {
		state = S_nonascii_text;
		*q++ = *p++;
		continue;
	    } else if (*p == '(') {
		state = S_dollar_paren;
		*q++ = *p++;
		continue;
	    } else {
		state = S_text;
	    }
	    break;

	case S_dollar_paren:
	    if (*p == 'C') {
		state = S_nonascii_text;
		*q++ = *p++;
		continue;
	    } else {
		state = S_text;
	    }
	    break;

	case S_paren:
	    if (*p == 'B' || *p == 'J' || *p == 'T') {
		state = S_text;
		*q++ = *p++;
		continue;
	    } else if (*p == 'I') {
		state = S_nonascii_text;
		*q++ = *p++;
		continue;
	    } else {
		state = S_text;
	    }
	    break;

	case S_nonascii_text:
	    if (*p == '\033') {
		if ((IS_CJK_TTY && !hidden) || stype != st_HTML) {
		    state = S_esc;
		    if (stype == st_URL) {
			REPLACE_STRING("%1B");
			p++;
			continue;
		    } else if (stype != st_HTML) {
			p++;
			continue;
		    }
		}
	    }
	    *q++ = *p++;
	    continue;

	case S_trans_byte:
	    /* character translation goes here */
	    /*
	     * Don't do anything if we have no string, or if original AND
	     * target character sets are both iso-8859-1, or if we are in CJK
	     * mode.
	     */
	    if (*p == '\0' || no_bytetrans) {
		state = S_got_outchar;
		break;
	    }

	    if (Back) {
		int rev_c;

		if ((*p) == HT_NON_BREAK_SPACE ||
		    (*p) == HT_EN_SPACE) {
		    if (plain_space) {
			code = *p = ' ';
			state = S_got_outchar;
			break;
		    } else {
			code = 160;
			if (LYCharSet_UC[cs_to].enc == UCT_ENC_8859 ||
			    (LYCharSet_UC[cs_to].like8859 & UCT_R_8859SPECL)) {
			    state = S_got_outchar;
			    break;
			} else if (!(LYCharSet_UC[cs_from].enc == UCT_ENC_8859
				     || (LYCharSet_UC[cs_from].like8859 & UCT_R_8859SPECL))) {
			    state = S_check_uni;
			    break;
			} else {
			    *(unsigned char *) p = UCH(160);
			}
		    }
		} else if ((*p) == LY_SOFT_HYPHEN) {
		    code = 173;
		    if (LYCharSet_UC[cs_to].enc == UCT_ENC_8859 ||
			(LYCharSet_UC[cs_to].like8859 & UCT_R_8859SPECL)) {
			state = S_got_outchar;
			break;
		    } else if (!(LYCharSet_UC[cs_from].enc == UCT_ENC_8859
				 || (LYCharSet_UC[cs_from].like8859 & UCT_R_8859SPECL))) {
			state = S_check_uni;
			break;
		    } else {
			*(unsigned char *) p = UCH(173);
		    }
#ifdef EXP_JAPANESEUTF8_SUPPORT
		} else if (output_utf8) {
		    if ((!strcmp(LYCharSet_UC[cs_from].MIMEname, "euc-jp") &&
			 (IS_EUC((unsigned char) (*p),
				 (unsigned char) (*(p + 1))))) ||
			(!strcmp(LYCharSet_UC[cs_from].MIMEname, "shift_jis") &&
			 (IS_SJIS_2BYTE((unsigned char) (*p),
					(unsigned char) (*(p + 1)))))) {
			code = UCTransJPToUni(p, 2, cs_from);
			p++;
			state = S_check_uni;
			break;
		    }
#endif
		} else if (code < 127 || T.transp) {
		    state = S_got_outchar;
		    break;
		}
		rev_c = UCReverseTransChar(*p, cs_to, cs_from);
		if (rev_c > 127) {
		    *p = (char) rev_c;
		    code = rev_c;
		    state = S_got_outchar;
		    break;
		}
	    } else if (code < 127) {
		state = S_got_outchar;
		break;
	    }

	    if (from_is_utf8) {
		if (((*p) & 0xc0) == 0xc0) {
		    puni = p;
		    code = UCGetUniFromUtf8String(&puni);
		    if (code <= 0) {
			code = UCH(*p);
		    } else {
			what = P_utf8;
		    }
		}
	    } else if (use_lynx_specials && !Back &&
		       (code == 160 || code == 173) &&
		       (LYCharSet_UC[cs_from].enc == UCT_ENC_8859 ||
			(LYCharSet_UC[cs_from].like8859 & UCT_R_8859SPECL))) {
		if (code == 160)
		    code = *p = HT_NON_BREAK_SPACE;
		else if (code == 173)
		    code = *p = LY_SOFT_HYPHEN;
		state = S_got_outchar;
		break;
	    } else if (T.trans_to_uni) {
		code = UCTransToUni(*p, cs_from);
		if (code <= 0) {
		    /* What else can we do? */
		    code = UCH(*p);
		}
	    } else if (!T.trans_from_uni) {
		state = S_got_outchar;
		break;
	    }
	    /*
	     * Substitute Lynx special character for 160 (nbsp) if
	     * use_lynx_specials is set.
	     */
	    if (use_lynx_specials && !Back &&
		(code == 160 || code == 173)) {
		code = ((code == 160 ? HT_NON_BREAK_SPACE : LY_SOFT_HYPHEN));
		state = S_got_outchar;
		break;
	    }

	    state = S_check_uni;
	    break;

	case S_check_ent:
	    if (*p == '&') {
		char *pp = p + 1;

		len = strlen(pp);
		/*
		 * Check for a numeric entity.  - FM
		 */
		if (*pp == '#' && len > 2 &&
		    (*(pp + 1) == 'x' || *(pp + 1) == 'X') &&
		    UCH(*(pp + 2)) < 127 &&
		    isxdigit(UCH(*(pp + 2)))) {
		    what = P_hex;
		    state = S_ncr;
		} else if (*pp == '#' && len > 2 &&
			   UCH(*(pp + 1)) < 127 &&
			   isdigit(UCH(*(pp + 1)))) {
		    what = P_decimal;
		    state = S_ncr;
		} else if (UCH(*pp) < 127 &&
			   isalpha(UCH(*pp))) {
		    what = P_named;
		    state = S_named;
		} else {
		    state = S_trans_byte;
		}
	    } else {
		state = S_trans_byte;
	    }
	    break;

	case S_ncr:
	    if (what == P_hex) {
		p += 3;
	    } else {		/* P_decimal */
		p += 2;
	    }
	    cp = p;
	    while (*p && UCH(*p) < 127 &&
		   (what == P_hex ? isxdigit(UCH(*p)) :
		    isdigit(UCH(*p)))) {
		p++;
	    }
	    /*
	     * Save the terminator and isolate the digit(s).  - FM
	     */
	    cpe = *p;
	    if (*p)
		*p++ = '\0';
	    /*
	     * Show the numeric entity if the value:
	     * (1) Is greater than 255 and unhandled Unicode.
	     * (2) Is less than 32, and not valid and we don't have HTCJK set.
	     * (3) Is 127 and we don't have HTPassHighCtrlRaw or HTCJK set.
	     * (4) Is 128 - 159 and we don't have HTPassHighCtrlNum set.
	     */
	    if ((((what == P_hex)
		  ? sscanf(cp, "%lx", &lcode)
		  : sscanf(cp, "%lu", &lcode)) != 1) ||
		lcode > 0x7fffffffL) {
		state = S_recover;
		break;
	    } else {
		code = LYcp1252ToUnicode((UCode_t) lcode);
		state = S_check_uni;
	    }
	    break;

	case S_check_uni:
	    /*
	     * Show the numeric entity if the value:
	     * (2) Is less than 32, and not valid and we don't have HTCJK set.
	     * (3) Is 127 and we don't have HTPassHighCtrlRaw or HTCJK set.
	     * (4) Is 128 - 159 and we don't have HTPassHighCtrlNum set.
	     */
	    if ((code < 32 &&
		 code != 9 && code != 10 && code != 13 &&
		 !IS_CJK_TTY) ||
		(code == 127 &&
		 !(HTPassHighCtrlRaw || IS_CJK_TTY)) ||
		(code > 127 && code < 160 &&
		 !HTPassHighCtrlNum)) {
		state = S_recover;
		break;
	    }
	    /*
	     * Convert the value as an unsigned char, hex escaped if isURL is
	     * set and it's 8-bit, and then recycle the terminator if it is not
	     * a semicolon.  - FM
	     */
	    if (code > 159 && stype == st_URL) {
		state = S_got_oututf8;
		break;
	    }
	    /*
	     * For 160 (nbsp), use that value if it's a hidden INPUT, otherwise
	     * use an ASCII space (32) if plain_space is TRUE, otherwise use
	     * the Lynx special character.  - FM
	     */
	    if (code == 160) {
		if (plain_space) {
		    code = ' ';
		    state = S_got_outchar;
		    break;
		} else if (use_lynx_specials) {
		    code = HT_NON_BREAK_SPACE;
		    state = S_got_outchar;
		    break;
		} else if ((hidden && !Back)
			   || (LYCharSet_UC[cs_to].codepoints & UCT_CP_SUPERSETOF_LAT1)
			   || LYCharSet_UC[cs_to].enc == UCT_ENC_8859
			   || (LYCharSet_UC[cs_to].like8859 &
			       UCT_R_8859SPECL)) {
		    state = S_got_outchar;
		    break;
		} else if (
			      (LYCharSet_UC[cs_to].repertoire & UCT_REP_SUPERSETOF_LAT1)) {
		    ;		/* nothing, may be translated later */
		} else {
		    code = ' ';
		    state = S_got_outchar;
		    break;
		}
	    }
	    /*
	     * For 173 (shy), use that value if it's a hidden INPUT, otherwise
	     * ignore it if plain_space is TRUE, otherwise use the Lynx special
	     * character.  - FM
	     */
	    if (code == 173) {
		if (plain_space) {
		    replace_buf[0] = '\0';
		    state = S_got_outstring;
		    break;
		} else if (Back &&
			   !(LYCharSet_UC[cs_to].enc == UCT_ENC_8859 ||
			     (LYCharSet_UC[cs_to].like8859 &
			      UCT_R_8859SPECL))) {
		    ;		/* nothing, may be translated later */
		} else if (hidden || Back) {
		    state = S_got_outchar;
		    break;
		} else if (use_lynx_specials) {
		    code = LY_SOFT_HYPHEN;
		    state = S_got_outchar;
		    break;
		}
	    }
	    /*
	     * Seek a translation from the chartrans tables.
	     */
	    if ((uck = UCTransUniChar(code,
				      cs_to)) >= 32 &&
		uck < 256 &&
		(uck < 127 || uck >= lowest_8)) {
		code = uck;
		state = S_got_outchar;
		break;
	    } else if ((uck == -4 ||
			(repl_translated_C0 &&
			 uck > 0 && uck < 32)) &&
		/*
		 * Not found; look for replacement string.
		 */
		       UCTransUniCharStr(replace_buf,
					 60, code,
					 cs_to,
					 0) >= 0) {
		state = S_got_outstring;
		break;
	    }
	    if (output_utf8 &&
		code > 127 && code < 0x7fffffffL) {
		state = S_got_oututf8;
		break;
	    }
	    /*
	     * For 8194 (ensp), 8195 (emsp), or 8201 (thinsp), use the
	     * character reference if it's a hidden INPUT, otherwise use an
	     * ASCII space (32) if plain_space is TRUE, otherwise use the Lynx
	     * special character.  - FM
	     */
	    if (code == 8194 || code == 8195 || code == 8201) {
		if (hidden) {
		    state = S_recover;
		} else if (plain_space) {
		    code = ' ';
		    state = S_got_outchar;
		} else {
		    code = HT_EN_SPACE;
		    state = S_got_outchar;
		}
		break;
		/*
		 * Ignore 8204 (zwnj), 8205 (zwj) 8206 (lrm), and 8207 (rlm),
		 * for now, if we got this far without finding a representation
		 * for them.
		 */
	    } else if (code == 8204 || code == 8205 ||
		       code == 8206 || code == 8207) {
		CTRACE((tfp, "LYUCFullyTranslateString: Ignoring '%"
			PRI_UCode_t "'.\n", code));
		replace_buf[0] = '\0';
		state = S_got_outstring;
		break;
		/*
		 * Show the numeric entity if the value:  (1) Is greater than
		 * 255 and unhandled Unicode.
		 */
	    } else if (code > 255) {
		/*
		 * Illegal or not yet handled value.  Return "&#" verbatim and
		 * continue from there.  - FM
		 */
		state = S_recover;
		break;
		/*
		 * If it's ASCII, or is 8-bit but HTPassEightBitNum is set or
		 * the character set is "ISO Latin 1", use it's value.  - FM
		 */
	    } else if (code < 161 ||
		       (code < 256 &&
			(HTPassEightBitNum || cs_to == LATIN1))) {
		/*
		 * No conversion needed.
		 */
		state = S_got_outchar;
		break;

		/* The following disabled section doesn't make sense any more. 
		 * It used to make sense in the past, when S_check_named would
		 * look in "old style" tables in addition to what it does now. 
		 * Disabling of going to S_check_name here prevents endless
		 * looping between S_check_uni and S_check_names states, which
		 * could occur here for Latin 1 codes for some cs_to if they
		 * had no translation in that cs_to.  Normally all cs_to
		 * *should* now have valid translations via UCTransUniChar or
		 * UCTransUniCharStr for all Latin 1 codes, so that we would
		 * not get here anyway, and no loop could occur.  Still, if we
		 * *do* get here, FALL THROUGH to case S_recover now.  - kw
		 */
#if 0
		/*
		 * If we get to here, convert and handle the character as a
		 * named entity.  - FM
		 */
	    } else {
		name = HTMLGetEntityName(code - 160);
		state = S_check_name;
		break;
#endif
	    }

	case S_recover:
	    if (what == P_decimal || what == P_hex) {
		/*
		 * Illegal or not yet handled value.  Return "&#" verbatim and
		 * continue from there.  - FM
		 */
		*q++ = '&';
		*q++ = '#';
		if (what == P_hex)
		    *q++ = 'x';
		if (cpe != '\0')
		    *(p - 1) = cpe;
		p = cp;
		state = S_done;
	    } else if (what == P_named) {
		*cp = cpe;
		*q++ = '&';
		state = S_done;
	    } else if (!T.output_utf8 && stype == st_HTML && !hidden &&
		       !(HTPassEightBitRaw &&
			 UCH(*p) >= lowest_8)) {
		sprintf(replace_buf, "U%.2" PRI_UCode_t "", code);

		state = S_got_outstring;
	    } else {
		puni = p;
		code = UCH(*p);
		state = S_got_outchar;
	    }
	    break;

	case S_named:
	    cp = ++p;
	    while (*cp && UCH(*cp) < 127 &&
		   isalnum(UCH(*cp)))
		cp++;
	    cpe = *cp;
	    *cp = '\0';
	    name = p;
	    state = S_check_name;
	    break;

	case S_check_name:
	    /*
	     * Seek the Unicode value for the named entity.
	     *
	     * !!!!  We manually recover the case of '=' terminator which is
	     * commonly found on query to CGI-scripts enclosed as href= URLs
	     * like "somepath/?x=1&yz=2" Without this dirty fix, submission of
	     * such URLs was broken if &yz string happened to be a recognized
	     * entity name.  - LP
	     */
	    if (((code = HTMLGetEntityUCValue(name)) > 0) &&
		!((cpe == '=') && (stype == st_URL))) {
		state = S_check_uni;
		break;
	    }
	    /*
	     * Didn't find the entity.  Return verbatim.
	     */
	    state = S_recover;
	    break;

	    /* * * O U T P U T   S T A T E S * * */

	case S_got_oututf8:
	    if (code > 255 ||
		(code >= 128 && LYCharSet_UC[cs_to].enc == UCT_ENC_UTF8)) {
		UCConvertUniToUtf8(code, replace_buf);
		state = S_got_outstring;
	    } else {
		state = S_got_outchar;
	    }
	    break;
	case S_got_outstring:
	    if (what == P_decimal || what == P_hex) {
		if (cpe != ';' && cpe != '\0')
		    *(--p) = cpe;
		p--;
	    } else if (what == P_named) {
		*cp = cpe;
		p = (*cp != ';') ? (cp - 1) : cp;
	    } else if (what == P_utf8) {
		p = puni;
	    }
	    if (replace_buf[0] == '\0') {
		state = S_next_char;
		break;
	    }
	    if (stype == st_URL) {
		code = replace_buf[0];	/* assume string OK if first char is */
		if (code >= 127 ||
		    (code < 32 && (code != 9 && code != 10 && code != 0))) {
		    state = S_put_urlstring;
		    break;
		}
	    }
	    REPLACE_STRING(replace_buf);
	    state = S_next_char;
	    break;
	case S_put_urlstring:
	    esc = HTEscape(replace_buf, URL_XALPHAS);
	    REPLACE_STRING(esc);
	    FREE(esc);
	    state = S_next_char;
	    break;
	case S_got_outchar:
	    if (what == P_decimal || what == P_hex) {
		if (cpe != ';' && cpe != '\0')
		    *(--p) = cpe;
		p--;
	    } else if (what == P_named) {
		*cp = cpe;
		p = (*cp != ';') ? (cp - 1) : cp;
	    } else if (what == P_utf8) {
		p = puni;
	    }
	    if (stype == st_URL &&
	    /*  Not a full HTEscape, only for 8bit and ctrl chars */
		(TOASCII(code) >= 127 ||	/* S/390 -- gil -- 1925 */
		 (code < ' ' && (code != '\t' && code != '\n')))) {
		state = S_put_urlchar;
		break;
	    } else if (!hidden && code == 10 && *p == 10
		       && q != qs && *(q - 1) == 13) {
		/*
		 * If this is not a hidden string, and the current char is the
		 * LF ('\n') of a CRLF pair, drop the CR ('\r').  - KW
		 */
		*(q - 1) = *p++;
		state = S_done;
		break;
	    }
	    *q++ = (char) code;
	    state = S_next_char;
	    break;
	case S_put_urlchar:
	    *q++ = '%';
	    REPLACE_CHAR(hex[(TOASCII(code) >> 4) & 15]);	/* S/390 -- gil -- 1944 */
	    REPLACE_CHAR(hex[(TOASCII(code) & 15)]);
	    /* fall through */
	case S_next_char:
	    p++;		/* fall through */
	case S_done:
	    state = S_text;
	    what = P_text;
	    /* for next round */
	}
    }

    *q = '\0';
    if (chunk) {
	HTChunkPutb(CHUNK, qs, (int) (q - qs + 1));	/* also terminates */
	if (stype == st_URL || stype == st_other) {
	    LYTrimHead(chunk->data);
	    LYTrimTail(chunk->data);
	}
	StrAllocCopy(*str, chunk->data);
	HTChunkFree(chunk);
    } else {
	if (stype == st_URL || stype == st_other) {
	    LYTrimHead(qs);
	    LYTrimTail(qs);
	}
    }
    return str;
}

#undef REPLACE_CHAR
#undef REPLACE_STRING

BOOL LYUCTranslateHTMLString(char **str,
			     int cs_from,
			     int cs_to,
			     int use_lynx_specials,
			     int plain_space,
			     int hidden,
			     CharUtil_st stype)
{
    BOOL ret = YES;

    /* May reallocate *str even if cs_to == 0 */
    if (!LYUCFullyTranslateString(str, cs_from, cs_to, TRUE,
				  use_lynx_specials, plain_space, hidden,
				  NO, stype)) {
	ret = NO;
    }
    return ret;
}

BOOL LYUCTranslateBackFormData(char **str,
			       int cs_from,
			       int cs_to,
			       int plain_space)
{
    char **ret;

    /* May reallocate *str */
    ret = (LYUCFullyTranslateString(str, cs_from, cs_to, FALSE,
				    NO, plain_space, YES,
				    YES, st_HTML));
    return (BOOL) (ret != NULL);
}

/*
 * Parse a parameter from an HTML META tag, i.e., the CONTENT.
 */
char *LYParseTagParam(char *from,
		      const char *name)
{
    size_t len = strlen(name);
    char *result = NULL;
    char *string = from;

    do {
	if ((string = strchr(string, ';')) == NULL)
	    return NULL;
	while (*string != '\0' && (*string == ';' || isspace(UCH(*string)))) {
	    string++;
	}
	if (strlen(string) < len)
	    return NULL;
    } while (strncasecomp(string, name, (int) len) != 0);
    string += len;
    while (*string != '\0' && (isspace(UCH(*string)) || *string == '=')) {
	string++;
    }

    StrAllocCopy(result, string);
    len = 0;
    while (isprint(UCH(string[len])) && !isspace(UCH(string[len]))) {
	len++;
    }
    result[len] = '\0';

    /*
     * Strip single quotes, just in case.
     */
    if (len > 2 && result[0] == '\'' && result[len - 1] == result[0]) {
	result[len - 1] = '\0';
	for (string = result; (string[0] = string[1]) != '\0'; ++string) ;
    }
    return result;
}

/*
 * Given a refresh-URL content string, parses the delay time and the URL
 * string.  Ignore the remainder of the content.
 */
void LYParseRefreshURL(char *content,
		       char **p_seconds,
		       char **p_address)
{
    char *cp;
    char *cp1 = NULL;
    char *Seconds = NULL;

    /*
     * Look for the Seconds field.  - FM
     */
    cp = LYSkipBlanks(content);
    if (*cp && isdigit(UCH(*cp))) {
	cp1 = cp;
	while (*cp1 && isdigit(UCH(*cp1)))
	    cp1++;
	StrnAllocCopy(Seconds, cp, (int) (cp1 - cp));
    }
    *p_seconds = Seconds;
    *p_address = LYParseTagParam(content, "URL");

    CTRACE((tfp,
	    "LYParseRefreshURL\n\tcontent: %s\n\tseconds: %s\n\taddress: %s\n",
	    content, NonNull(*p_seconds), NonNull(*p_address)));
}

/*
 *  This function processes META tags in HTML streams. - FM
 */
void LYHandleMETA(HTStructured * me, const BOOL *present,
		  const char **value,
		  char **include GCC_UNUSED)
{
    char *http_equiv = NULL, *name = NULL, *content = NULL;
    char *href = NULL, *id_string = NULL, *temp = NULL;
    char *cp, *cp0, *cp1 = NULL;
    int url_type = 0;

    if (!me || !present)
	return;

    /*
     * Load the attributes for possible use by Lynx.  - FM
     */
    if (present[HTML_META_HTTP_EQUIV] &&
	non_empty(value[HTML_META_HTTP_EQUIV])) {
	StrAllocCopy(http_equiv, value[HTML_META_HTTP_EQUIV]);
	convert_to_spaces(http_equiv, TRUE);
	LYUCTranslateHTMLString(&http_equiv, me->tag_charset, me->tag_charset,
				NO, NO, YES, st_other);
	if (*http_equiv == '\0') {
	    FREE(http_equiv);
	}
    }
    if (present[HTML_META_NAME] &&
	non_empty(value[HTML_META_NAME])) {
	StrAllocCopy(name, value[HTML_META_NAME]);
	convert_to_spaces(name, TRUE);
	LYUCTranslateHTMLString(&name, me->tag_charset, me->tag_charset,
				NO, NO, YES, st_other);
	if (*name == '\0') {
	    FREE(name);
	}
    }
    if (present[HTML_META_CONTENT] &&
	non_empty(value[HTML_META_CONTENT])) {
	/*
	 * Technically, we should be creating a comma-separated list, but META
	 * tags come one at a time, and we'll handle (or ignore) them as each
	 * is received.  Also, at this point, we only trim leading and trailing
	 * blanks from the CONTENT value, without translating any named
	 * entities or numeric character references, because how we should do
	 * that depends on what type of information it contains, and whether or
	 * not any of it might be sent to the screen.  - FM
	 */
	StrAllocCopy(content, value[HTML_META_CONTENT]);
	convert_to_spaces(content, FALSE);
	LYTrimHead(content);
	LYTrimTail(content);
	if (*content == '\0') {
	    FREE(content);
	}
    }
    CTRACE((tfp,
	    "LYHandleMETA: HTTP-EQUIV=\"%s\" NAME=\"%s\" CONTENT=\"%s\"\n",
	    NONNULL(http_equiv),
	    NONNULL(name),
	    NONNULL(content)));

    /*
     * Make sure we have META name/value pairs to handle.  - FM
     */
    if (!(http_equiv || name) || !content)
	goto free_META_copies;

    /*
     * Check for a no-cache Pragma
     * or Cache-Control directive. - FM
     */
    if (!strcasecomp(NonNull(http_equiv), "Pragma") ||
	!strcasecomp(NonNull(http_equiv), "Cache-Control")) {
	LYUCTranslateHTMLString(&content, me->tag_charset, me->tag_charset,
				NO, NO, YES, st_other);
	if (!strcasecomp(content, "no-cache")) {
	    me->node_anchor->no_cache = TRUE;
	    HText_setNoCache(me->text);
	}

	/*
	 * If we didn't get a Cache-Control MIME header, and the META has one,
	 * convert to lowercase, store it in the anchor element, and if we
	 * haven't yet set no_cache, check whether we should.  - FM
	 */
	if ((!me->node_anchor->cache_control) &&
	    !strcasecomp(NonNull(http_equiv), "Cache-Control")) {
	    LYLowerCase(content);
	    StrAllocCopy(me->node_anchor->cache_control, content);
	    if (me->node_anchor->no_cache == FALSE) {
		cp0 = content;
		while ((cp = strstr(cp0, "no-cache")) != NULL) {
		    cp += 8;
		    while (*cp != '\0' && WHITE(*cp))
			cp++;
		    if (*cp == '\0' || *cp == ';') {
			me->node_anchor->no_cache = TRUE;
			HText_setNoCache(me->text);
			break;
		    }
		    cp0 = cp;
		}
		if (me->node_anchor->no_cache == TRUE)
		    goto free_META_copies;
		cp0 = content;
		while ((cp = strstr(cp0, "max-age")) != NULL) {
		    cp += 7;
		    while (*cp != '\0' && WHITE(*cp))
			cp++;
		    if (*cp == '=') {
			cp++;
			while (*cp != '\0' && WHITE(*cp))
			    cp++;
			if (isdigit(UCH(*cp))) {
			    cp0 = cp;
			    while (isdigit(UCH(*cp)))
				cp++;
			    if (*cp0 == '0' && cp == (cp0 + 1)) {
				me->node_anchor->no_cache = TRUE;
				HText_setNoCache(me->text);
				break;
			    }
			}
		    }
		    cp0 = cp;
		}
	    }
	}

	/*
	 * Check for an Expires directive. - FM
	 */
    } else if (!strcasecomp(NonNull(http_equiv), "Expires")) {
	/*
	 * If we didn't get an Expires MIME header, store it in the anchor
	 * element, and if we haven't yet set no_cache, check whether we
	 * should.  Note that we don't accept a Date header via META tags,
	 * because it's likely to be untrustworthy, but do check for a Date
	 * header from a server when making the comparison.  - FM
	 */
	LYUCTranslateHTMLString(&content, me->tag_charset, me->tag_charset,
				NO, NO, YES, st_other);
	StrAllocCopy(me->node_anchor->expires, content);
	if (me->node_anchor->no_cache == FALSE) {
	    if (!strcmp(content, "0")) {
		/*
		 * The value is zero, which we treat as an absolute no-cache
		 * directive.  - FM
		 */
		me->node_anchor->no_cache = TRUE;
		HText_setNoCache(me->text);
	    } else if (me->node_anchor->date != NULL) {
		/*
		 * We have a Date header, so check if the value is less than or
		 * equal to that.  - FM
		 */
		if (LYmktime(content, TRUE) <=
		    LYmktime(me->node_anchor->date, TRUE)) {
		    me->node_anchor->no_cache = TRUE;
		    HText_setNoCache(me->text);
		}
	    } else if (LYmktime(content, FALSE) == 0) {
		/*
		 * We don't have a Date header, and the value is in past for
		 * us.  - FM
		 */
		me->node_anchor->no_cache = TRUE;
		HText_setNoCache(me->text);
	    }
	}

	/*
	 * Check for a text/html Content-Type with a charset directive, if we
	 * didn't already set the charset via a server's header.  - AAC & FM
	 */
    } else if (isEmpty(me->node_anchor->charset) &&
	       !strcasecomp(NonNull(http_equiv), "Content-Type")) {
	LYUCcharset *p_in = NULL;
	LYUCcharset *p_out = NULL;

	LYUCTranslateHTMLString(&content, me->tag_charset, me->tag_charset,
				NO, NO, YES, st_other);
	LYLowerCase(content);

	if ((cp1 = strstr(content, "charset")) != NULL) {
	    BOOL chartrans_ok = NO;
	    char *cp3 = NULL, *cp4;
	    int chndl;

	    cp1 += 7;
	    while (*cp1 == ' ' || *cp1 == '=' || *cp1 == '"')
		cp1++;

	    StrAllocCopy(cp3, cp1);	/* copy to mutilate more */
	    for (cp4 = cp3; (*cp4 != '\0' && *cp4 != '"' &&
			     *cp4 != ';' && *cp4 != ':' &&
			     !WHITE(*cp4)); cp4++) {
		;		/* do nothing */
	    }
	    *cp4 = '\0';
	    cp4 = cp3;
	    chndl = UCGetLYhndl_byMIME(cp3);

#ifdef CAN_SWITCH_DISPLAY_CHARSET
	    /* Allow a switch to a more suitable display charset */
	    if (Switch_Display_Charset(chndl, SWITCH_DISPLAY_CHARSET_MAYBE)) {
		/* UCT_STAGE_STRUCTURED and UCT_STAGE_HTEXT
		   should have the same setting for UCInfoStage. */
		HTAnchor_getUCInfoStage(me->node_anchor, UCT_STAGE_STRUCTURED);

		me->outUCLYhndl = current_char_set;
		HTAnchor_setUCInfoStage(me->node_anchor,
					current_char_set,
					UCT_STAGE_HTEXT,
					UCT_SETBY_MIME);	/* highest priorty! */
		HTAnchor_setUCInfoStage(me->node_anchor,
					current_char_set,
					UCT_STAGE_STRUCTURED,
					UCT_SETBY_MIME);	/* highest priorty! */
		me->outUCI = HTAnchor_getUCInfoStage(me->node_anchor,
						     UCT_STAGE_HTEXT);
		/* The SGML stage will be reset in change_chartrans_handling */
	    }
#endif

	    if (UCCanTranslateFromTo(chndl, current_char_set)) {
		chartrans_ok = YES;
		StrAllocCopy(me->node_anchor->charset, cp4);
		HTAnchor_setUCInfoStage(me->node_anchor, chndl,
					UCT_STAGE_PARSER,
					UCT_SETBY_STRUCTURED);
	    } else if (chndl < 0) {
		/*
		 * Got something but we don't recognize it.
		 */
		chndl = UCLYhndl_for_unrec;
		if (chndl < 0)	/* UCLYhndl_for_unrec not defined :-( */
		    chndl = UCLYhndl_for_unspec;	/* always >= 0 */
		if (UCCanTranslateFromTo(chndl, current_char_set)) {
		    chartrans_ok = YES;
		    HTAnchor_setUCInfoStage(me->node_anchor, chndl,
					    UCT_STAGE_PARSER,
					    UCT_SETBY_STRUCTURED);
		}
	    }
	    if (chartrans_ok) {
		p_in = HTAnchor_getUCInfoStage(me->node_anchor,
					       UCT_STAGE_PARSER);
		p_out = HTAnchor_setUCInfoStage(me->node_anchor,
						current_char_set,
						UCT_STAGE_HTEXT,
						UCT_SETBY_DEFAULT);
		if (!p_out) {
		    /*
		     * Try again.
		     */
		    p_out = HTAnchor_getUCInfoStage(me->node_anchor,
						    UCT_STAGE_HTEXT);
		}
		if (!strcmp(p_in->MIMEname, "x-transparent")) {
		    HTPassEightBitRaw = TRUE;
		    HTAnchor_setUCInfoStage(me->node_anchor,
					    HTAnchor_getUCLYhndl(me->node_anchor,
								 UCT_STAGE_HTEXT),
					    UCT_STAGE_PARSER,
					    UCT_SETBY_DEFAULT);
		}
		if (!strcmp(p_out->MIMEname, "x-transparent")) {
		    HTPassEightBitRaw = TRUE;
		    HTAnchor_setUCInfoStage(me->node_anchor,
					    HTAnchor_getUCLYhndl(me->node_anchor,
								 UCT_STAGE_PARSER),
					    UCT_STAGE_HTEXT,
					    UCT_SETBY_DEFAULT);
		}
		if ((p_in->enc != UCT_ENC_CJK)
#ifdef EXP_JAPANESEUTF8_SUPPORT
		    && (p_in->enc != UCT_ENC_UTF8)
#endif
		    ) {
		    HTCJK = NOCJK;
		    if (!(p_in->codepoints &
			  UCT_CP_SUBSETOF_LAT1) &&
			chndl == current_char_set) {
			HTPassEightBitRaw = TRUE;
		    }
		} else if (p_out->enc == UCT_ENC_CJK) {
		    Set_HTCJK(p_in->MIMEname, p_out->MIMEname);
		}
		LYGetChartransInfo(me);
		/*
		 * Update the chartrans info homologously to a Content-Type
		 * MIME header with a charset parameter.  - FM
		 */
		if (me->UCLYhndl != chndl) {
		    HTAnchor_setUCInfoStage(me->node_anchor, chndl,
					    UCT_STAGE_MIME,
					    UCT_SETBY_STRUCTURED);
		    HTAnchor_setUCInfoStage(me->node_anchor, chndl,
					    UCT_STAGE_PARSER,
					    UCT_SETBY_STRUCTURED);
		    me->inUCLYhndl = HTAnchor_getUCLYhndl(me->node_anchor,
							  UCT_STAGE_PARSER);
		    me->inUCI = HTAnchor_getUCInfoStage(me->node_anchor,
							UCT_STAGE_PARSER);
		}
		UCSetTransParams(&me->T,
				 me->inUCLYhndl, me->inUCI,
				 me->outUCLYhndl, me->outUCI);
	    } else {
		/*
		 * Cannot translate.  If according to some heuristic the given
		 * charset and the current display character both are likely to
		 * be like ISO-8859 in structure, pretend we have some kind of
		 * match.
		 */
		BOOL given_is_8859 = (BOOL) (!StrNCmp(cp4, "iso-8859-", 9) &&
					     isdigit(UCH(cp4[9])));
		BOOL given_is_8859like = (BOOL) (given_is_8859
						 || !StrNCmp(cp4, "windows-", 8)
						 || !StrNCmp(cp4, "cp12", 4)
						 || !StrNCmp(cp4, "cp-12", 5));
		BOOL given_and_display_8859like = (BOOL) (given_is_8859like &&
							  (strstr(LYchar_set_names[current_char_set],
								  "ISO-8859") ||
							   strstr(LYchar_set_names[current_char_set],
								  "windows-")));

		if (given_is_8859) {
		    cp1 = &cp4[10];
		    while (*cp1 &&
			   isdigit(UCH((*cp1))))
			cp1++;
		    *cp1 = '\0';
		}
		if (given_and_display_8859like) {
		    StrAllocCopy(me->node_anchor->charset, cp4);
		    HTPassEightBitRaw = TRUE;
		}
		HTAlert(*cp4 ? cp4 : me->node_anchor->charset);

	    }
	    FREE(cp3);

	    if (me->node_anchor->charset) {
		CTRACE((tfp,
			"LYHandleMETA: New charset: %s\n",
			me->node_anchor->charset));
	    }
	}
	/*
	 * Set the kcode element based on the charset.  - FM
	 */
	HText_setKcode(me->text, me->node_anchor->charset, p_in);

	/*
	 * Check for a Refresh directive.  - FM
	 */
    } else if (!strcasecomp(NonNull(http_equiv), "Refresh")) {
	char *Seconds = NULL;

	LYUCTranslateHTMLString(&content, me->tag_charset, me->tag_charset,
				NO, NO, YES, st_other);
	LYParseRefreshURL(content, &Seconds, &href);

	if (Seconds) {
	    if (href) {
		/*
		 * We found a URL field, so check it out.  - FM
		 */
		if (!LYLegitimizeHREF(me, &href, TRUE, FALSE)) {
		    /*
		     * The specs require a complete URL, but this is a
		     * Netscapism, so don't expect the author to know that.  -
		     * FM
		     */
		    HTUserMsg(REFRESH_URL_NOT_ABSOLUTE);
		    /*
		     * Use the document's address as the base.  - FM
		     */
		    if (*href != '\0') {
			temp = HTParse(href,
				       me->node_anchor->address, PARSE_ALL);
			StrAllocCopy(href, temp);
			FREE(temp);
		    } else {
			StrAllocCopy(href, me->node_anchor->address);
			HText_setNoCache(me->text);
		    }

		} else {
		    /*
		     * Check whether to fill in localhost.  - FM
		     */
		    LYFillLocalFileURL(&href,
				       (me->inBASE ?
					me->base_href : me->node_anchor->address));
		}

		/*
		 * Set the no_cache flag if the Refresh URL is the same as the
		 * document's address.  - FM
		 */
		if (!strcmp(href, me->node_anchor->address)) {
		    HText_setNoCache(me->text);
		}
	    } else {
		/*
		 * We didn't find a URL field, so use the document's own
		 * address and set the no_cache flag.  - FM
		 */
		StrAllocCopy(href, me->node_anchor->address);
		HText_setNoCache(me->text);
	    }
	    /*
	     * Check for an anchor in http or https URLs.  - FM
	     */
	    cp = NULL;
#ifndef DONT_TRACK_INTERNAL_LINKS
	    /* id_string seems to be used wrong below if given.
	       not that it matters much.  avoid setting it here. - kw */
	    if ((StrNCmp(href, "http", 4) == 0) &&
		(cp = strchr(href, '#')) != NULL) {
		StrAllocCopy(id_string, cp);
		*cp = '\0';
	    }
#endif
	    if (me->inA) {
		/*
		 * Ugh!  The META tag, which is a HEAD element, is in an
		 * Anchor, which is BODY element.  All we can do is close the
		 * Anchor and cross our fingers.  - FM
		 */
		if (me->inBoldA == TRUE && me->inBoldH == FALSE)
		    HText_appendCharacter(me->text, LY_BOLD_END_CHAR);
		me->inBoldA = FALSE;
		HText_endAnchor(me->text, me->CurrentANum);
		me->inA = FALSE;
		me->CurrentANum = 0;
	    }
	    me->CurrentA = HTAnchor_findChildAndLink
		(
		    me->node_anchor,	/* Parent */
		    id_string,	/* Tag */
		    href,	/* Addresss */
		    (HTLinkType *) 0);	/* Type */
	    if (id_string)
		*cp = '#';
	    FREE(id_string);
	    LYEnsureSingleSpace(me);
	    if (me->inUnderline == FALSE)
		HText_appendCharacter(me->text, LY_UNDERLINE_START_CHAR);
	    HTML_put_string(me, "REFRESH(");
	    HTML_put_string(me, Seconds);
	    HTML_put_string(me, " sec):");
	    FREE(Seconds);
	    if (me->inUnderline == FALSE)
		HText_appendCharacter(me->text, LY_UNDERLINE_END_CHAR);
	    HTML_put_character(me, ' ');
	    me->in_word = NO;
	    HText_beginAnchor(me->text, me->inUnderline, me->CurrentA);
	    if (me->inBoldH == FALSE)
		HText_appendCharacter(me->text, LY_BOLD_START_CHAR);
	    HTML_put_string(me, href);
	    FREE(href);
	    if (me->inBoldH == FALSE)
		HText_appendCharacter(me->text, LY_BOLD_END_CHAR);
	    HText_endAnchor(me->text, 0);
	    LYEnsureSingleSpace(me);
	}

	/*
	 * Check for a suggested filename via a Content-Disposition with a
	 * filename=name.suffix in it, if we don't already have it via a server
	 * header.  - FM
	 */
    } else if (isEmpty(me->node_anchor->SugFname) &&
	       !strcasecomp((http_equiv ?
			     http_equiv : ""), "Content-Disposition")) {
	cp = content;
	while (*cp != '\0' && strncasecomp(cp, "filename", 8))
	    cp++;
	if (*cp != '\0') {
	    cp = LYSkipBlanks(cp + 8);
	    if (*cp == '=')
		cp++;
	    cp = LYSkipBlanks(cp);
	    if (*cp != '\0') {
		StrAllocCopy(me->node_anchor->SugFname, cp);
		if (*me->node_anchor->SugFname == '"') {
		    if ((cp = strchr((me->node_anchor->SugFname + 1),
				     '"')) != NULL) {
			*(cp + 1) = '\0';
			HTMIME_TrimDoubleQuotes(me->node_anchor->SugFname);
			if (isEmpty(me->node_anchor->SugFname)) {
			    FREE(me->node_anchor->SugFname);
			}
		    } else {
			FREE(me->node_anchor->SugFname);
		    }
		}
#if defined(UNIX) && !defined(DOSPATH)
		/*
		 * If blanks are not legal for local filenames, replace them
		 * with underscores.
		 */
		if ((cp = me->node_anchor->SugFname) != NULL) {
		    while (*cp != '\0') {
			if (isspace(UCH(*cp)))
			    *cp = '_';
			++cp;
		    }
		}
#endif
	    }
	}
	/*
	 * Check for a Set-Cookie directive.  - AK
	 */
    } else if (!strcasecomp(NonNull(http_equiv), "Set-Cookie")) {
	/*
	 * This will need to be updated when Set-Cookie/Set-Cookie2 handling is
	 * finalized.  For now, we'll still assume "historical" cookies in META
	 * directives.  - FM
	 */
	url_type = is_url(me->inBASE ?
			  me->base_href : me->node_anchor->address);
	if (url_type == HTTP_URL_TYPE || url_type == HTTPS_URL_TYPE) {
	    LYSetCookie(content,
			NULL,
			(me->inBASE ?
			 me->base_href : me->node_anchor->address));
	}
    }

    /*
     * Free the copies.  - FM
     */
  free_META_copies:
    FREE(http_equiv);
    FREE(name);
    FREE(content);
}

/*
 *  This function handles P elements in HTML streams.
 *  If start is TRUE it handles a start tag, and if
 *  FALSE, an end tag.	We presently handle start
 *  and end tags identically, but this can lead to
 *  a different number of blank lines between the
 *  current paragraph and subsequent text when a P
 *  end tag is present or not in the markup. - FM
 */
void LYHandlePlike(HTStructured * me, const BOOL *present,
		   const char **value,
		   char **include GCC_UNUSED,
		   int align_idx,
		   int start)
{
    if (TRUE) {
	/*
	 * FIG content should be a true block, which like P inherits the
	 * current style.  APPLET is like character elements or an ALT
	 * attribute, unless it content contains a block element.  If we
	 * encounter a P in either's content, we set flags to treat the content
	 * as a block.  - FM
	 */
	if (start) {
	    if (me->inFIG)
		me->inFIGwithP = TRUE;

	    if (me->inAPPLET)
		me->inAPPLETwithP = TRUE;
	}

	UPDATE_STYLE;
	if (me->List_Nesting_Level >= 0) {
	    /*
	     * We're in a list.  Treat P as an instruction to create one blank
	     * line, if not already present, then fall through to handle
	     * attributes, with the "second line" margins.  - FM
	     */
	    if (me->inP) {
		if (me->inFIG || me->inAPPLET ||
		    me->inCAPTION || me->inCREDIT ||
		    me->sp->style->spaceAfter > 0 ||
		    (start && me->sp->style->spaceBefore > 0)) {
		    LYEnsureDoubleSpace(me);
		} else {
		    LYEnsureSingleSpace(me);
		}
	    }
	} else if (me->sp[0].tag_number == HTML_ADDRESS) {
	    /*
	     * We're in an ADDRESS.  Treat P as an instruction to start a
	     * newline, if needed, then fall through to handle attributes.  -
	     * FM
	     */
	    if (!HText_LastLineEmpty(me->text, FALSE)) {
		HText_setLastChar(me->text, ' ');	/* absorb white space */
		HText_appendCharacter(me->text, '\r');
	    }
	} else {
	    if (start) {
		if (!(me->inLABEL && !me->inP)) {
		    HText_appendParagraph(me->text);
		}
	    } else if (me->sp->style->spaceAfter > 0) {
		LYEnsureDoubleSpace(me);
	    } else {
		LYEnsureSingleSpace(me);
	    }
	    me->inLABEL = FALSE;
	}
	me->in_word = NO;

	if (LYoverride_default_alignment(me)) {
	    me->sp->style->alignment = LYstyles(me->sp[0].tag_number)->alignment;
	} else if ((me->List_Nesting_Level >= 0 &&
		    (me->sp->style->id == ST_DivCenter ||
		     me->sp->style->id == ST_DivLeft ||
		     me->sp->style->id == ST_DivRight)) ||
		   ((me->Division_Level < 0) &&
		    (me->sp->style->id == ST_Normal ||
		     me->sp->style->id == ST_Preformatted))) {
	    me->sp->style->alignment = HT_LEFT;
	} else {
	    me->sp->style->alignment = (short) me->current_default_alignment;
	}

	if (start) {
	    if (present && present[align_idx] && value[align_idx]) {
		if (!strcasecomp(value[align_idx], "center") &&
		    !(me->List_Nesting_Level >= 0 && !me->inP))
		    me->sp->style->alignment = HT_CENTER;
		else if (!strcasecomp(value[align_idx], "right") &&
			 !(me->List_Nesting_Level >= 0 && !me->inP))
		    me->sp->style->alignment = HT_RIGHT;
		else if (!strcasecomp(value[align_idx], "left") ||
			 !strcasecomp(value[align_idx], "justify"))
		    me->sp->style->alignment = HT_LEFT;
	    }

	}

	/*
	 * Mark that we are starting a new paragraph and don't have any of it's
	 * text yet.  - FM
	 */
	me->inP = FALSE;
    }

    return;
}

/*
 *  This function handles SELECT elements in HTML streams.
 *  If start is TRUE it handles a start tag, and if FALSE,
 *  an end tag. - FM
 */
void LYHandleSELECT(HTStructured * me, const BOOL *present,
		    const char **value,
		    char **include GCC_UNUSED,
		    int start)
{
    int i;

    if (start == TRUE) {
	char *name = NULL;
	BOOLEAN multiple = NO;
	char *size = NULL;

	/*
	 * Initialize the disable attribute.
	 */
	me->select_disabled = FALSE;

	/*
	 * Check for unclosed TEXTAREA.
	 */
	if (me->inTEXTAREA) {
	    if (LYBadHTML(me)) {
		LYShowBadHTML("Bad HTML: Missing TEXTAREA end tag\n");
	    }
	}

	/*
	 * Set to know we are in a select tag.
	 */
	me->inSELECT = TRUE;

	if (!(present && present[HTML_SELECT_NAME] &&
	      non_empty(value[HTML_SELECT_NAME]))) {
	    StrAllocCopy(name, "");
	} else if (strchr(value[HTML_SELECT_NAME], '&') == NULL) {
	    StrAllocCopy(name, value[HTML_SELECT_NAME]);
	} else {
	    StrAllocCopy(name, value[HTML_SELECT_NAME]);
	    UNESCAPE_FIELDNAME_TO_STD(&name);
	}
	if (present && present[HTML_SELECT_MULTIPLE])
	    multiple = YES;
	if (present && present[HTML_SELECT_DISABLED])
	    me->select_disabled = TRUE;
	if (present && present[HTML_SELECT_SIZE] &&
	    non_empty(value[HTML_SELECT_SIZE])) {
	    /*
	     * Let the size be determined by the number of OPTIONs.  - FM
	     */
	    CTRACE((tfp, "LYHandleSELECT: Ignoring SIZE=\"%s\" for SELECT.\n",
		    value[HTML_SELECT_SIZE]));
	}

	if (me->inBoldH == TRUE &&
	    (multiple == NO || LYSelectPopups == FALSE)) {
	    HText_appendCharacter(me->text, LY_BOLD_END_CHAR);
	    me->inBoldH = FALSE;
	    me->needBoldH = TRUE;
	}
	if (me->inUnderline == TRUE &&
	    (multiple == NO || LYSelectPopups == FALSE)) {
	    HText_appendCharacter(me->text, LY_UNDERLINE_END_CHAR);
	    me->inUnderline = FALSE;
	}

	if ((multiple == NO && LYSelectPopups == TRUE) &&
	    (me->sp[0].tag_number == HTML_PRE || me->inPRE == TRUE ||
	     !me->sp->style->freeFormat) &&
	    HText_LastLineSize(me->text, FALSE) > (LYcolLimit - 7)) {
	    /*
	     * Force a newline when we're using a popup in a PRE block and are
	     * within 7 columns from the right margin.  This will allow for the
	     * '[' popup designator and help avoid a wrap in the underscore
	     * placeholder for the retracted popup entry in the HText
	     * structure.  - FM
	     */
	    HTML_put_character(me, '\n');
	    me->in_word = NO;
	}

	LYCheckForID(me, present, value, (int) HTML_SELECT_ID);

	HText_beginSelect(name, ATTR_CS_IN, multiple, size);
	FREE(name);
	FREE(size);

	me->first_option = TRUE;
    } else {
	/*
	 * Handle end tag.
	 */
	char *ptr;

	/*
	 * Make sure we had a select start tag.
	 */
	if (!me->inSELECT) {
	    if (LYBadHTML(me)) {
		LYShowBadHTML("Bad HTML: Unmatched SELECT end tag\n");
	    }
	    return;
	}

	/*
	 * Set to know that we are no longer in a select tag.
	 */
	me->inSELECT = FALSE;

	/*
	 * Clear the disable attribute.
	 */
	me->select_disabled = FALSE;

	/*
	 * Finish the data off.
	 */
	HTChunkTerminate(&me->option);
	/*
	 * Finish the previous option.
	 */
	ptr = HText_setLastOptionValue(me->text,
				       me->option.data,
				       me->LastOptionValue,
				       LAST_ORDER,
				       me->LastOptionChecked,
				       me->UCLYhndl,
				       ATTR_CS_IN);
	FREE(me->LastOptionValue);

	me->LastOptionChecked = FALSE;

	if (HTCurSelectGroupType == F_CHECKBOX_TYPE ||
	    LYSelectPopups == FALSE) {
	    /*
	     * Start a newline after the last checkbox/button option.
	     */
	    LYEnsureSingleSpace(me);
	} else {
	    /*
	     * Output popup box with the default option to screen, but use
	     * non-breaking spaces for output.
	     */
	    if (ptr &&
		me->sp[0].tag_number == HTML_PRE && strlen(ptr) > 6) {
		/*
		 * The code inadequately handles OPTION fields in PRE tags. 
		 * We'll put up a minimum of 6 characters, and if any more
		 * would exceed the wrap column, we'll ignore them.
		 */
		for (i = 0; i < 6; i++) {
		    if (*ptr == ' ')
			HText_appendCharacter(me->text, HT_NON_BREAK_SPACE);
		    else
			HText_appendCharacter(me->text, *ptr);
		    ptr++;
		}
	    }
	    for (; non_empty(ptr); ptr++) {
		if (*ptr == ' ')
		    HText_appendCharacter(me->text, HT_NON_BREAK_SPACE);
		else
		    HText_appendCharacter(me->text, *ptr);
	    }
	    /*
	     * Add end option character.
	     */
	    if (!me->first_option) {
		HText_appendCharacter(me->text, ']');
		HText_setLastChar(me->text, ']');
		me->in_word = YES;
	    }
	}
	HTChunkClear(&me->option);

	if (me->Underline_Level > 0 && me->inUnderline == FALSE) {
	    HText_appendCharacter(me->text, LY_UNDERLINE_START_CHAR);
	    me->inUnderline = TRUE;
	}
	if (me->needBoldH == TRUE && me->inBoldH == FALSE) {
	    HText_appendCharacter(me->text, LY_BOLD_START_CHAR);
	    me->inBoldH = TRUE;
	    me->needBoldH = FALSE;
	}
    }
}

/*
 *  This function strips white characters and
 *  generally fixes up attribute values that
 *  were received from the SGML parser and
 *  are to be treated as partial or absolute
 *  URLs. - FM
 */
int LYLegitimizeHREF(HTStructured * me, char **href,
		     int force_slash,
		     int strip_dots)
{
    int url_type = 0;
    char *p = NULL;
    char *pound = NULL;
    const char *Base = NULL;

    if (!me || !href || isEmpty(*href))
	return (url_type);

    if (!LYTrimStartfile(*href)) {
	/*
	 * Collapse spaces in the actual URL, but just protect against tabs or
	 * newlines in the fragment, if present.  This seeks to cope with
	 * atrocities inflicted on the Web by authoring tools such as
	 * Frontpage.  - FM
	 */

	/*  Before working on spaces check if we have any, usually none. */
	p = LYSkipNonBlanks(*href);

	if (*p) {		/* p == first space character */
	    /* no reallocs below, all converted in place */

	    pound = findPoundSelector(*href);

	    if (pound != NULL && pound < p) {
		convert_to_spaces(p, FALSE);	/* done */

	    } else {
		if (pound != NULL)
		    *pound = '\0';	/* mark */

		/*
		 * No blanks really belong in the HREF,
		 * but if it refers to an actual file,
		 * it may actually have blanks in the name.
		 * Try to accommodate. See also HTParse().
		 */
		if (LYRemoveNewlines(p) || strchr(p, '\t') != 0) {
		    LYRemoveBlanks(p);	/* a compromise... */
		}

		if (pound != NULL) {
		    p = strchr(p, '\0');
		    *pound = '#';	/* restore */
		    convert_to_spaces(pound, FALSE);
		    if (p < pound)
			strcpy(p, pound);
		}
	    }
	}
    }
    if (**href == '\0')
	return (url_type);

    TRANSLATE_AND_UNESCAPE_TO_STD(href);

    Base = me->inBASE ?
	me->base_href : me->node_anchor->address;

    url_type = is_url(*href);
    if (!url_type && force_slash && **href == '.' &&
	(!strcmp(*href, ".") || !strcmp(*href, "..")) &&
	!isFILE_URL(Base)) {
	/*
	 * The Fielding RFC/ID for resolving partial HREFs says that a slash
	 * should be on the end of the preceding symbolic element for "." and
	 * "..", but all tested browsers only do that for an explicit "./" or
	 * "../", so we'll respect the RFC/ID only if force_slash was TRUE and
	 * it's not a file URL.  - FM
	 */
	StrAllocCat(*href, "/");
    }
    if ((!url_type && LYStripDotDotURLs && strip_dots && **href == '.') &&
	!strncasecomp(Base, "http", 4)) {
	/*
	 * We will be resolving a partial reference versus an http or https
	 * URL, and it has lead dots, which may be retained when resolving via
	 * HTParse(), but the request would fail if the first element of the
	 * resultant path is two dots, because no http or https server accepts
	 * such paths, and the current URL draft, likely to become an RFC, says
	 * that it's optional for the UA to strip them as a form of error
	 * recovery.  So we will, recursively, for http/https URLs, like the
	 * "major market browsers" which made this problem so common on the
	 * Web, but we'll also issue a message about it, such that the bad
	 * partial reference might get corrected by the document provider.  -
	 * FM
	 */
	char *temp = NULL, *path = NULL, *cp;
	const char *str = "";

	temp = HTParse(*href, Base, PARSE_ALL);
	path = HTParse(temp, "", PARSE_PATH + PARSE_PUNCTUATION);
	if (!StrNCmp(path, "/..", 3)) {
	    cp = (path + 3);
	    if (LYIsHtmlSep(*cp) || *cp == '\0') {
		if (Base[4] == 's') {
		    str = "s";
		}
		CTRACE((tfp,
			"LYLegitimizeHREF: Bad value '%s' for http%s URL.\n",
			*href, str));
		CTRACE((tfp, "                  Stripping lead dots.\n"));
		if (!me->inBadHREF) {
		    HTUserMsg(BAD_PARTIAL_REFERENCE);
		    me->inBadHREF = TRUE;
		}
	    }
	    if (*cp == '\0') {
		StrAllocCopy(*href, "/");
	    } else if (LYIsHtmlSep(*cp)) {
		while (!StrNCmp(cp, "/..", 3)) {
		    if (*(cp + 3) == '/') {
			cp += 3;
			continue;
		    } else if (*(cp + 3) == '\0') {
			*(cp + 1) = '\0';
			*(cp + 2) = '\0';
		    }
		    break;
		}
		StrAllocCopy(*href, cp);
	    }
	}
	FREE(temp);
	FREE(path);
    }
    return (url_type);
}

/*
 *  This function checks for a Content-Base header,
 *  and if not present, a Content-Location header
 *  which is an absolute URL, and sets the BASE
 *  accordingly.  If set, it will be replaced by
 *  any BASE tag in the HTML stream, itself. - FM
 */
void LYCheckForContentBase(HTStructured * me)
{
    char *cp = NULL;
    BOOL present[HTML_BASE_ATTRIBUTES];
    const char *value[HTML_BASE_ATTRIBUTES];
    int i;

    if (!(me && me->node_anchor))
	return;

    if (me->node_anchor->content_base != NULL) {
	/*
	 * We have a Content-Base value.  Use it if it's non-zero length.  - FM
	 */
	if (*me->node_anchor->content_base == '\0')
	    return;
	StrAllocCopy(cp, me->node_anchor->content_base);
	LYRemoveBlanks(cp);
    } else if (me->node_anchor->content_location != NULL) {
	/*
	 * We didn't have a Content-Base value, but do have a Content-Location
	 * value.  Use it if it's an absolute URL.  - FM
	 */
	if (*me->node_anchor->content_location == '\0')
	    return;
	StrAllocCopy(cp, me->node_anchor->content_location);
	LYRemoveBlanks(cp);
	if (!is_url(cp)) {
	    FREE(cp);
	    return;
	}
    } else {
	/*
	 * We had neither a Content-Base nor Content-Location value.  - FM
	 */
	return;
    }

    /*
     * If we collapsed to a zero-length value, ignore it.  - FM
     */
    if (*cp == '\0') {
	FREE(cp);
	return;
    }

    /*
     * Pass the value to HTML_start_element as the HREF of a BASE tag.  - FM
     */
    for (i = 0; i < HTML_BASE_ATTRIBUTES; i++)
	present[i] = NO;
    present[HTML_BASE_HREF] = YES;
    value[HTML_BASE_HREF] = (const char *) cp;
    (*me->isa->start_element) (me, HTML_BASE, present, value,
			       0, 0);
    FREE(cp);
}

/*
 *  This function creates NAMEd Anchors if a non-zero-length NAME
 *  or ID attribute was present in the tag. - FM
 */
void LYCheckForID(HTStructured * me, const BOOL *present,
		  const char **value,
		  int attribute)
{
    HTChildAnchor *ID_A = NULL;
    char *temp = NULL;

    if (!(me && me->text))
	return;

    if (present && present[attribute]
	&& non_empty(value[attribute])) {
	/*
	 * Translate any named or numeric character references.  - FM
	 */
	StrAllocCopy(temp, value[attribute]);
	LYUCTranslateHTMLString(&temp, me->tag_charset, me->tag_charset,
				NO, NO, YES, st_URL);

	/*
	 * Create the link if we still have a non-zero-length string.  - FM
	 */
	if ((temp[0] != '\0') &&
	    (ID_A = HTAnchor_findChildAndLink
	     (
		 me->node_anchor,	/* Parent */
		 temp,		/* Tag */
		 NULL,		/* Addresss */
		 (HTLinkType *) 0))) {	/* Type */
	    HText_beginAnchor(me->text, me->inUnderline, ID_A);
	    HText_endAnchor(me->text, 0);
	}
	FREE(temp);
    }
}

/*
 *  This function creates a NAMEd Anchor for the ID string
 *  passed to it directly as an argument.  It assumes the
 *  does not need checking for character references. - FM
 */
void LYHandleID(HTStructured * me, const char *id)
{
    HTChildAnchor *ID_A = NULL;

    if (!(me && me->text) ||
	isEmpty(id))
	return;

    /*
     * Create the link if we still have a non-zero-length string.  - FM
     */
    if ((ID_A = HTAnchor_findChildAndLink
	 (
	     me->node_anchor,	/* Parent */
	     id,		/* Tag */
	     NULL,		/* Addresss */
	     (HTLinkType *) 0)) != NULL) {	/* Type */
	HText_beginAnchor(me->text, me->inUnderline, ID_A);
	HText_endAnchor(me->text, 0);
    }
}

/*
 *  This function checks whether we want to override
 *  the current default alignment for paragraphs and
 *  instead use that specified in the element's style
 *  sheet. - FM
 */
BOOLEAN LYoverride_default_alignment(HTStructured * me)
{
    if (!me)
	return NO;

    switch (me->sp[0].tag_number) {
    case HTML_BLOCKQUOTE:
    case HTML_BQ:
    case HTML_NOTE:
    case HTML_FN:
    case HTML_ADDRESS:
	me->sp->style->alignment = HT_LEFT;
	return YES;

    default:
	break;
    }
    return NO;
}

/*
 *  This function inserts newlines if needed to create double spacing,
 *  and sets the left margin for subsequent text to the second line
 *  indentation of the current style. - FM
 */
void LYEnsureDoubleSpace(HTStructured * me)
{
    if (!me || !me->text)
	return;

    if (!HText_LastLineEmpty(me->text, FALSE)) {
	HText_setLastChar(me->text, ' ');	/* absorb white space */
	HText_appendCharacter(me->text, '\r');
	HText_appendCharacter(me->text, '\r');
    } else if (!HText_PreviousLineEmpty(me->text, FALSE)) {
	HText_setLastChar(me->text, ' ');	/* absorb white space */
	HText_appendCharacter(me->text, '\r');
    } else if (me->List_Nesting_Level >= 0) {
	HText_NegateLineOne(me->text);
    }
    me->in_word = NO;
    return;
}

/*
 *  This function inserts a newline if needed to create single spacing,
 *  and sets the left margin for subsequent text to the second line
 *  indentation of the current style. - FM
 */
void LYEnsureSingleSpace(HTStructured * me)
{
    if (!me || !me->text)
	return;

    if (!HText_LastLineEmpty(me->text, FALSE)) {
	HText_setLastChar(me->text, ' ');	/* absorb white space */
	HText_appendCharacter(me->text, '\r');
    } else if (me->List_Nesting_Level >= 0) {
	HText_NegateLineOne(me->text);
    }
    me->in_word = NO;
    return;
}

/*
 *  This function resets paragraph alignments for block
 *  elements which do not have a defined style sheet. - FM
 */
void LYResetParagraphAlignment(HTStructured * me)
{
    if (!me)
	return;

    if (me->List_Nesting_Level >= 0 ||
	((me->Division_Level < 0) &&
	 (me->sp->style->id == ST_Normal ||
	  me->sp->style->id == ST_Preformatted))) {
	me->sp->style->alignment = HT_LEFT;
    } else {
	me->sp->style->alignment = (short) me->current_default_alignment;
    }
    return;
}

/*
 *  This example function checks whether the given anchor has
 *  an address with a file scheme, and if so, loads it into the
 *  the SGML parser's context->url element, which was passed as
 *  the second argument.  The handle_comment() calling function in
 *  SGML.c then calls LYDoCSI() in LYUtils.c to insert HTML markup
 *  into the corresponding stream, homologously to an SSI by an
 *  HTTP server. - FM
 *
 *  For functions similar to this but which depend on details of
 *  the HTML handler's internal data, the calling interface should
 *  be changed, and functions in SGML.c would have to make sure not
 *  to call such functions inappropriately (e.g., calling a function
 *  specific to the Lynx_HTML_Handler when SGML.c output goes to
 *  some other HTStructured object like in HTMLGen.c), or the new
 *  functions could be added to the SGML.h interface.
 */
BOOLEAN LYCheckForCSI(HTParentAnchor *anchor,
		      char **url)
{
    if (!(anchor && anchor->address))
	return FALSE;

    if (!isFILE_URL(anchor->address))
	return FALSE;

    if (!LYisLocalHost(anchor->address))
	return FALSE;

    StrAllocCopy(*url, anchor->address);
    return TRUE;
}

/*
 *  This function is called from the SGML parser to look at comments
 *  and see whether we should collect some info from them.  Currently
 *  it only looks for comments with Message-Id and Subject info, in the
 *  exact form generated by MHonArc for archived mailing list.  If found,
 *  the info is stored in the document's HTParentAnchor.  It can later be
 *  used for generating a mail response.
 *
 *  We are extra picky here because there isn't any official definition
 *  for these kinds of comments - we might (and still can) misinterpret
 *  arbitrary comments as something they aren't.
 *
 *  If something doesn't look right, for example invalid characters, the
 *  strings are not stored.  Mail responses will use something else as
 *  the subject, probably the document URL, and will not have an
 *  In-Reply-To header.
 *
 *  All this is a hack - to do this the right way, mailing list archivers
 *  would have to agree on some better mechanism to make this kind of info
 *  from original mail headers available, for example using LINK.  - kw
 */
BOOLEAN LYCommentHacks(HTParentAnchor *anchor,
		       const char *comment)
{
    const char *cp;
    size_t len;

    if (comment == NULL)
	return FALSE;

    if (!(anchor && anchor->address))
	return FALSE;

    if (StrNCmp(comment, "!--X-Message-Id: ", 17) == 0) {
	char *messageid = NULL;
	char *p;

	for (cp = comment + 17; *cp; cp++) {
	    if (UCH(*cp) >= 127 || !isgraph(UCH(*cp))) {
		break;
	    }
	}
	if (strcmp(cp, " --")) {
	    return FALSE;
	}
	cp = comment + 17;
	StrAllocCopy(messageid, cp);
	/* This should be ok - message-id should only contain 7-bit ASCII */
	if (!LYUCTranslateHTMLString(&messageid, 0, 0, NO, NO, YES, st_URL))
	    return FALSE;
	for (p = messageid; *p; p++) {
	    if (UCH(*p) >= 127 || !isgraph(UCH(*p))) {
		break;
	    }
	}
	if (strcmp(p, " --")) {
	    FREE(messageid);
	    return FALSE;
	}
	if ((p = strchr(messageid, '@')) == NULL || p[1] == '\0') {
	    FREE(messageid);
	    return FALSE;
	}
	p = messageid;
	if ((len = strlen(p)) >= 8 && !strcmp(&p[len - 3], " --")) {
	    p[len - 3] = '\0';
	} else {
	    FREE(messageid);
	    return FALSE;
	}
	if (HTAnchor_setMessageID(anchor, messageid)) {
	    FREE(messageid);
	    return TRUE;
	} else {
	    FREE(messageid);
	    return FALSE;
	}
    }
    if (StrNCmp(comment, "!--X-Subject: ", 14) == 0) {
	char *subject = NULL;
	char *p;

	for (cp = comment + 14; *cp; cp++) {
	    if (UCH(*cp) >= 127 || !isprint(UCH(*cp))) {
		return FALSE;
	    }
	}
	cp = comment + 14;
	StrAllocCopy(subject, cp);
	/* @@@
	 * This may not be the right thing for the subject - but mail
	 * subjects shouldn't contain 8-bit characters in raw form anyway.
	 * We have to unescape character entities, since that's what MHonArc
	 * seems to generate.  But if after that there are 8-bit characters
	 * the string is rejected.  We would probably not know correctly
	 * what charset to assume anyway - the mail sender's can differ from
	 * the archive's.  And the code for sending mail cannot deal well
	 * with 8-bit characters - we should not put them in the Subject
	 * header in raw form, but don't have MIME encoding implemented.
	 * Someone may want to do more about this...  - kw
	 */
	if (!LYUCTranslateHTMLString(&subject, 0, 0, NO, YES, NO, st_HTML))
	    return FALSE;
	for (p = subject; *p; p++) {
	    if (UCH(*p) >= 127 || !isprint(UCH(*p))) {
		FREE(subject);
		return FALSE;
	    }
	}
	p = subject;
	if ((len = strlen(p)) >= 4 && !strcmp(&p[len - 3], " --")) {
	    p[len - 3] = '\0';
	} else {
	    FREE(subject);
	    return FALSE;
	}
	if (HTAnchor_setSubject(anchor, subject)) {
	    FREE(subject);
	    return TRUE;
	} else {
	    FREE(subject);
	    return FALSE;
	}
    }

    return FALSE;
}

    /*
     * Create the Title with any left-angle-brackets converted to &lt; entities
     * and any ampersands converted to &amp; entities.  - FM
     *
     * Convert 8-bit letters to &#xUUUU to avoid dependencies from display
     * character set which may need changing.  Do NOT convert any 8-bit chars
     * if we have CJK display.  - LP
     */
void LYformTitle(char **dst,
		 const char *src)
{
    if (HTCJK == JAPANESE) {
	char *tmp_buffer = NULL;

	if ((tmp_buffer = (char *) malloc(strlen(src) + 1)) == 0)
	    outofmem(__FILE__, "LYformTitle");

	assert(tmp_buffer != NULL);

	switch (kanji_code) {	/* 1997/11/22 (Sat) 09:28:00 */
	case EUC:
	    TO_EUC((const unsigned char *) src, (unsigned char *) tmp_buffer);
	    break;
	case SJIS:
	    TO_SJIS((const unsigned char *) src, (unsigned char *) tmp_buffer);
	    break;
	default:
	    CTRACE((tfp, "\nLYformTitle: kanji_code is an unexpected value."));
	    strcpy(tmp_buffer, src);
	    break;
	}
	StrAllocCopy(*dst, tmp_buffer);
	FREE(tmp_buffer);
    } else {
	StrAllocCopy(*dst, src);
    }
}