1 /* -*- mode: C; c-file-style: "gnu"; indent-tabs-mode: nil; -*- */
3 /* GLIB - Library of useful routines for C programming
4 * Copyright (C) 2008 Red Hat, Inc.
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General
17 * Public License along with this library; if not, write to the
18 * Free Software Foundation, Inc., 59 Temple Place, Suite 330,
19 * Boston, MA 02111-1307, USA.
26 #include "ghostutils.h"
30 #include "gstrfuncs.h"
36 * @short_description: Internet hostname utilities
38 * Functions for manipulating internet hostnames; in particular, for
39 * converting between Unicode and ASCII-encoded forms of
40 * Internationalized Domain Names (IDNs).
43 * url="http://www.ietf.org/rfc/rfc3490.txt">Internationalized Domain
44 * Names for Applications (IDNA)</ulink> standards allow for the use
45 * of Unicode domain names in applications, while providing
46 * backward-compatibility with the old ASCII-only DNS, by defining an
47 * ASCII-Compatible Encoding of any given Unicode name, which can be
48 * used with non-IDN-aware applications and protocols. (For example,
49 * "Παν語.org" maps to "xn--4wa8awb4637h.org".)
52 #define IDNA_ACE_PREFIX "xn--"
53 #define IDNA_ACE_PREFIX_LEN 4
55 /* Punycode constants, from RFC 3492. */
57 #define PUNYCODE_BASE 36
58 #define PUNYCODE_TMIN 1
59 #define PUNYCODE_TMAX 26
60 #define PUNYCODE_SKEW 38
61 #define PUNYCODE_DAMP 700
62 #define PUNYCODE_INITIAL_BIAS 72
63 #define PUNYCODE_INITIAL_N 0x80
65 #define PUNYCODE_IS_BASIC(cp) ((guint)(cp) < 0x80)
67 /* Encode/decode a single base-36 digit */
69 encode_digit (guint dig)
74 return dig - 26 + '0';
78 decode_digit (gchar dig)
80 if (dig >= 'A' && dig <= 'Z')
82 else if (dig >= 'a' && dig <= 'z')
84 else if (dig >= '0' && dig <= '9')
85 return dig - '0' + 26;
90 /* Punycode bias adaptation algorithm, RFC 3492 section 6.1 */
98 delta = firsttime ? delta / PUNYCODE_DAMP : delta / 2;
99 delta += delta / numpoints;
102 while (delta > ((PUNYCODE_BASE - PUNYCODE_TMIN) * PUNYCODE_TMAX) / 2)
104 delta /= PUNYCODE_BASE - PUNYCODE_TMIN;
108 return k + ((PUNYCODE_BASE - PUNYCODE_TMIN + 1) * delta /
109 (delta + PUNYCODE_SKEW));
112 /* Punycode encoder, RFC 3492 section 6.3. The algorithm is
113 * sufficiently bizarre that it's not really worth trying to explain
117 punycode_encode (const gchar *input_utf8,
118 gsize input_utf8_length,
121 guint delta, handled_chars, num_basic_chars, bias, j, q, k, t, digit;
122 gunichar n, m, *input;
124 gboolean success = FALSE;
126 /* Convert from UTF-8 to Unicode code points */
127 input = g_utf8_to_ucs4 (input_utf8, input_utf8_length, NULL,
128 &input_length, NULL);
132 /* Copy basic chars */
133 for (j = num_basic_chars = 0; j < input_length; j++)
135 if (PUNYCODE_IS_BASIC (input[j]))
137 g_string_append_c (output, g_ascii_tolower (input[j]));
142 g_string_append_c (output, '-');
144 handled_chars = num_basic_chars;
146 /* Encode non-basic chars */
148 bias = PUNYCODE_INITIAL_BIAS;
149 n = PUNYCODE_INITIAL_N;
150 while (handled_chars < input_length)
152 /* let m = the minimum {non-basic} code point >= n in the input */
153 for (m = G_MAXUINT, j = 0; j < input_length; j++)
155 if (input[j] >= n && input[j] < m)
159 if (m - n > (G_MAXUINT - delta) / (handled_chars + 1))
161 delta += (m - n) * (handled_chars + 1);
164 for (j = 0; j < input_length; j++)
171 else if (input[j] == n)
174 for (k = PUNYCODE_BASE; ; k += PUNYCODE_BASE)
178 else if (k >= bias + PUNYCODE_TMAX)
184 digit = t + (q - t) % (PUNYCODE_BASE - t);
185 g_string_append_c (output, encode_digit (digit));
186 q = (q - t) / (PUNYCODE_BASE - t);
189 g_string_append_c (output, encode_digit (q));
190 bias = adapt (delta, handled_chars + 1, handled_chars == num_basic_chars);
207 /* From RFC 3454, Table B.1 */
208 #define idna_is_junk(ch) ((ch) == 0x00AD || (ch) == 0x1806 || (ch) == 0x200B || (ch) == 0x2060 || (ch) == 0xFEFF || (ch) == 0x034F || (ch) == 0x180B || (ch) == 0x180C || (ch) == 0x180D || (ch) == 0x200C || (ch) == 0x200D || ((ch) >= 0xFE00 && (ch) <= 0xFE0F))
210 /* Scan @str for "junk" and return a cleaned-up string if any junk
211 * is found. Else return %NULL.
214 remove_junk (const gchar *str,
217 GString *cleaned = NULL;
221 for (p = str; len == -1 ? *p : p < str + len; p = g_utf8_next_char (p))
223 ch = g_utf8_get_char (p);
224 if (idna_is_junk (ch))
228 cleaned = g_string_new (NULL);
229 g_string_append_len (cleaned, str, p - str);
233 g_string_append_unichar (cleaned, ch);
237 return g_string_free (cleaned, FALSE);
242 static inline gboolean
243 contains_uppercase_letters (const gchar *str,
248 for (p = str; len == -1 ? *p : p < str + len; p = g_utf8_next_char (p))
250 if (g_unichar_isupper (g_utf8_get_char (p)))
256 static inline gboolean
257 contains_non_ascii (const gchar *str,
262 for (p = str; len == -1 ? *p : p < str + len; p++)
264 if ((guchar)*p > 0x80)
270 /* RFC 3454, Appendix C. ish. */
271 static inline gboolean
272 idna_is_prohibited (gunichar ch)
274 switch (g_unichar_type (ch))
276 case G_UNICODE_CONTROL:
277 case G_UNICODE_FORMAT:
278 case G_UNICODE_UNASSIGNED:
279 case G_UNICODE_PRIVATE_USE:
280 case G_UNICODE_SURROGATE:
281 case G_UNICODE_LINE_SEPARATOR:
282 case G_UNICODE_PARAGRAPH_SEPARATOR:
283 case G_UNICODE_SPACE_SEPARATOR:
286 case G_UNICODE_OTHER_SYMBOL:
287 if (ch == 0xFFFC || ch == 0xFFFD ||
288 (ch >= 0x2FF0 && ch <= 0x2FFB))
292 case G_UNICODE_NON_SPACING_MARK:
293 if (ch == 0x0340 || ch == 0x0341)
302 /* RFC 3491 IDN cleanup algorithm. */
304 nameprep (const gchar *hostname,
307 gchar *name, *tmp = NULL, *p;
309 /* It would be nice if we could do this without repeatedly
310 * allocating strings and converting back and forth between
311 * gunichars and UTF-8... The code does at least avoid doing most of
312 * the sub-operations when they would just be equivalent to a
316 /* Remove presentation-only characters */
317 name = remove_junk (hostname, len);
324 name = (gchar *)hostname;
326 /* Convert to lowercase */
327 if (contains_uppercase_letters (name, len))
329 name = g_utf8_strdown (name, len);
335 /* If there are no UTF8 characters, we're done. */
336 if (!contains_non_ascii (name, len))
338 if (name == (gchar *)hostname)
339 return len == -1 ? g_strdup (hostname) : g_strndup (hostname, len);
345 name = g_utf8_normalize (name, len, G_NORMALIZE_NFKC);
352 /* KC normalization may have created more capital letters (eg,
353 * angstrom -> capital A with ring). So we have to lowercasify a
354 * second time. (This is more-or-less how the nameprep algorithm
355 * does it. If tolower(nfkc(tolower(X))) is guaranteed to be the
356 * same as tolower(nfkc(X)), then we could skip the first tolower,
357 * but I'm not sure it is.)
359 if (contains_uppercase_letters (name, -1))
361 name = g_utf8_strdown (name, -1);
366 /* Check for prohibited characters */
367 for (p = name; *p; p = g_utf8_next_char (p))
369 if (idna_is_prohibited (g_utf8_get_char (p)))
377 /* FIXME: We're supposed to verify certain constraints on bidi
378 * characters, but glib does not appear to have that information.
386 * g_hostname_to_ascii:
387 * @hostname: a valid UTF-8 or ASCII hostname
389 * Converts @hostname to its canonical ASCII form; an ASCII-only
390 * string containing no uppercase letters and not ending with a
393 * Return value: an ASCII hostname, which must be freed, or %NULL if
394 * @hostname is in some way invalid.
399 g_hostname_to_ascii (const gchar *hostname)
401 gchar *name, *label, *p;
406 label = name = nameprep (hostname, -1);
410 out = g_string_new (NULL);
415 for (p = label; *p && *p != '.'; p++)
417 if ((guchar)*p > 0x80)
425 if (!strncmp (label, IDNA_ACE_PREFIX, IDNA_ACE_PREFIX_LEN))
428 g_string_append (out, IDNA_ACE_PREFIX);
429 if (!punycode_encode (label, llen, out))
433 g_string_append_len (out, label, llen);
435 if (out->len - oldlen > 63)
439 if (*label && *++label)
440 g_string_append_c (out, '.');
445 return g_string_free (out, FALSE);
449 g_string_free (out, TRUE);
454 * g_hostname_is_non_ascii:
455 * @hostname: a hostname
457 * Tests if @hostname contains Unicode characters. If this returns
458 * %TRUE, you need to encode the hostname with g_hostname_to_ascii()
459 * before using it in non-IDN-aware contexts.
461 * Note that a hostname might contain a mix of encoded and unencoded
462 * segments, and so it is possible for g_hostname_is_non_ascii() and
463 * g_hostname_is_ascii_encoded() to both return %TRUE for a name.
465 * Return value: %TRUE if @hostname contains any non-ASCII characters
470 g_hostname_is_non_ascii (const gchar *hostname)
472 return contains_non_ascii (hostname, -1);
475 /* Punycode decoder, RFC 3492 section 6.2. As with punycode_encode(),
476 * read the RFC if you want to understand what this is actually doing.
479 punycode_decode (const gchar *input,
483 GArray *output_chars;
486 guint oldi, w, k, digit, t;
489 n = PUNYCODE_INITIAL_N;
491 bias = PUNYCODE_INITIAL_BIAS;
493 split = input + input_length - 1;
494 while (split > input && *split != '-')
498 output_chars = g_array_sized_new (FALSE, FALSE, sizeof (gunichar),
500 input_length -= (split - input) + 1;
501 while (input < split)
503 gunichar ch = (gunichar)*input++;
504 if (!PUNYCODE_IS_BASIC (ch))
506 g_array_append_val (output_chars, ch);
511 output_chars = g_array_new (FALSE, FALSE, sizeof (gunichar));
517 for (k = PUNYCODE_BASE; ; k += PUNYCODE_BASE)
521 digit = decode_digit (*input++);
522 if (digit >= PUNYCODE_BASE)
524 if (digit > (G_MAXUINT - i) / w)
529 else if (k >= bias + PUNYCODE_TMAX)
535 if (w > G_MAXUINT / (PUNYCODE_BASE - t))
537 w *= (PUNYCODE_BASE - t);
540 bias = adapt (i - oldi, output_chars->len + 1, oldi == 0);
542 if (i / (output_chars->len + 1) > G_MAXUINT - n)
544 n += i / (output_chars->len + 1);
545 i %= (output_chars->len + 1);
547 g_array_insert_val (output_chars, i++, n);
550 for (i = 0; i < output_chars->len; i++)
551 g_string_append_unichar (output, g_array_index (output_chars, gunichar, i));
552 g_array_free (output_chars, TRUE);
556 g_array_free (output_chars, TRUE);
561 * g_hostname_to_unicode:
562 * @hostname: a valid UTF-8 or ASCII hostname
564 * Converts @hostname to its canonical presentation form; a UTF-8
565 * string in Unicode normalization form C, containing no uppercase
566 * letters, no forbidden characters, and no ASCII-encoded segments,
567 * and not ending with a trailing dot.
569 * Of course if @hostname is not an internationalized hostname, then
570 * the canonical presentation form will be entirely ASCII.
572 * Return value: a UTF-8 hostname, which must be freed, or %NULL if
573 * @hostname is in some way invalid.
578 g_hostname_to_unicode (const gchar *hostname)
583 out = g_string_new (NULL);
587 llen = strcspn (hostname, ".");
588 if (!g_ascii_strncasecmp (hostname, IDNA_ACE_PREFIX, IDNA_ACE_PREFIX_LEN))
590 hostname += IDNA_ACE_PREFIX_LEN;
591 llen -= IDNA_ACE_PREFIX_LEN;
592 if (!punycode_decode (hostname, llen, out))
594 g_string_free (out, TRUE);
600 gchar *canonicalized = nameprep (hostname, llen);
604 g_string_free (out, TRUE);
607 g_string_append (out, canonicalized);
608 g_free (canonicalized);
612 if (*hostname && *++hostname)
613 g_string_append_c (out, '.');
617 return g_string_free (out, FALSE);
621 * g_hostname_is_ascii_encoded:
622 * @hostname: a hostname
624 * Tests if @hostname contains segments with an ASCII-compatible
625 * encoding of an Internationalized Domain Name. If this returns
626 * %TRUE, you should decode the hostname with g_hostname_to_unicode()
627 * before displaying it to the user.
629 * Note that a hostname might contain a mix of encoded and unencoded
630 * segments, and so it is possible for g_hostname_is_non_ascii() and
631 * g_hostname_is_ascii_encoded() to both return %TRUE for a name.
633 * Return value: %TRUE if @hostname contains any ASCII-encoded
639 g_hostname_is_ascii_encoded (const gchar *hostname)
643 if (!g_ascii_strncasecmp (hostname, IDNA_ACE_PREFIX, IDNA_ACE_PREFIX_LEN))
645 hostname = strchr (hostname, '.');
652 * g_hostname_is_ip_address:
653 * @hostname: a hostname (or IP address in string form)
655 * Tests if @hostname is the string form of an IPv4 or IPv6 address.
656 * (Eg, "192.168.0.1".)
658 * Return value: %TRUE if @hostname is an IP address
663 g_hostname_is_ip_address (const gchar *hostname)
666 gint nsegments, octet;
668 /* On Linux we could implement this using inet_pton, but the Windows
669 * equivalent of that requires linking against winsock, so we just
670 * figure this out ourselves. Tested by tests/hostutils.c.
673 p = (char *)hostname;
679 /* If it contains a ':', it's an IPv6 address (assuming it's an
680 * IP address at all). This consists of eight ':'-separated
681 * segments, each containing a 1-4 digit hex number, except that
682 * optionally: (a) the last two segments can be replaced by an
683 * IPv4 address, and (b) a single span of 1 to 8 "0000" segments
684 * can be replaced with just "::".
689 while (*p && nsegments < 8)
691 /* Each segment after the first must be preceded by a ':'.
692 * (We also handle half of the "string starts with ::" case
695 if (p != (char *)hostname || (p[0] == ':' && p[1] == ':'))
702 /* If there's another ':', it means we're skipping some segments */
703 if (*p == ':' && !skipped)
708 /* Handle the "string ends with ::" case */
715 /* Read the segment, make sure it's valid. */
716 for (end = p; g_ascii_isxdigit (*end); end++)
718 if (end == p || end > p + 4)
723 if ((nsegments == 6 && !skipped) || (nsegments <= 6 && skipped))
733 return !*p && (nsegments == 8 || skipped);
738 /* Parse IPv4: N.N.N.N, where each N <= 255 and doesn't have leading 0s. */
739 for (nsegments = 0; nsegments < 4; nsegments++)
748 /* Check the segment; a little tricker than the IPv6 case since
749 * we can't allow extra leading 0s, and we can't assume that all
750 * strings of valid length are within range.
757 for (end = p; g_ascii_isdigit (*end); end++)
758 octet = 10 * octet + (*end - '0');
760 if (end == p || end > p + 3 || octet > 255)
766 /* If there's nothing left to parse, then it's ok. */