libmdn, libmdnlite - Mulitilingual Domain Name Handling Libraries

#include <mdn/api.h>

mdn_result_t mdn_nameinit(void)

mdn_result_t mdn_encodename(int actions, const char *from, char *to, size_t tolen)

mdn_result_t mdn_decodename(int actions, const char *from, char *to, size_t tolen)

mdn_result_t mdn_enable(int on_off)

char * mdn_result_tostring(mdn_result_t result)

The libmdn and libmdnlite libraries support various manipulations of multilingual domain names, including:

-

encoding convesion

-

name preparation

They are designed according to IDNA framework where each application must do necessary preparations for the multilingual domain names before passing them to the resolver.

To help applications do the preparation, the libraries provide easy-to-use, high-level interface for the work.

Both libraries provide almost the same API. The difference between them is that libmdn internally uses iconv function to provide encoding conversion from UTF-8 to the local encoding (such as iso-8859-1, usually determined by the current locale), and vise versa. libmdnlite is lightweight version of libmdn. It assumes local encoding is UTF-8 so that it never uses iconv.

This manual describes only a small subset of the API that the libraries provide, most important functions for application programmers. For other API, please refer to the mDNkit's specification document (which is not yet available) or the header files typically found under `/usr/local/include/mdn/' on your system.

The mdn_nameinit function initializes the whole library, and reads the default configuration file mdn.conf, which includes configuration parameters for multilingual domain name preparation. If mdn_nameinit is called more than once, the library initialization will take place only at the first call while the configuration file will be (re)read at every call.

If there are no errors, mdn_nameinit returns mdn_success. Otherwise, the returned value indicates the cause of the error. See the section ``RETURN VALUES'' below for the error codes.

Usually you don't have to call this function explicitly because it is implicitly called when mdn_encodename or mdn_decodename is first called without prior calling of mdn_nameinit.

mdn_encodename function performs name preparation and encoding conversion on the multilingual domain name specified by from, and stores the result to to, whose length is specified by tolen. actions is a bitwise-OR of the following macros, specifying which subprocesses in the encoding process are to be employed.

MDN_LOCALCONV     Local encoding <-> UTF-8 conversion
MDN_DELIMMAP      Delimiter mapping
MDN_LOCALMAP      Local mapping
MDN_NAMEPREP      NAMEPREP
                  (inculding unassigned codepoint check)
MDN_IDNCONV       UTF-8 <-> ACE conversion

Details of this encoding process can be found in the section ``NAME ENCODING''.

For ordinary application, also the MDN_ENCODE_APP macro is provided for the convenience. It is eqivalent to

MDN_LOCALCONV|MDN_DELIMMAP|MDN_LOCALMAP|MDN_NAMEPREP|MDN_IDNCONV

in libmdn, and eqivalent to

MDN_DELIMMAP|MDN_LOCALMAP|MDN_NAMEPREP|MDN_IDNCONV

in libmdnlite.

mdn_decodename function performs the reverse of mdn_encodename. It converts the multilingual domain name given by from, which is represented in a special encoding called ACE, to the application's local codeset and stores into to, whose length is specified by tolen. actions specifies which subprocesses in decoding process take place, as in mdn_encodename, except that only CWMDN_IDNCONV, CWMDN_NAMEPREP and CWMDN_LOCALCONV are allowed. Details of this decoding process can be found in the section ``NAME DECODING''.

For ordinary application, also the MDN_DECODE_APP macro is provided for the convenience. It is eqivalent to

MDN_IDNCONV|MDN_NAMEPREP|MDN_LOCALCONV

in libmdn, and eqivalent to

MDN_IDNCONV|MDN_NAMEPREP

in libmdnlite.

All of the functions above return error code of type mdn_result_t. All codes other than mdn_success indicates some kind of failure. mdn_result_tostring function takes an error code result and returns a pointer to the corresponding message string.

Name encoding is a process that transforms the specified multilingual domain name to a certain string suitable for name resolution. The name encoding process consists of the following steps.

(1) Encoding Conversion (local codeset to UCS)

Convert the encoding of the given domain name to Unicode/ISO10646 UTF-8 format string.

This step is performed if actions given to mdn_encodename contains MDN_LOCALCONV or equals to MDN_ENCODE_APP. Note that libmdnlite doesn't support this step.

The source encoding must be one of the two encodings below:

-

The application's local encoding (codeset).

-

ACE specified by the configuration file. The latter is suppored because name decoding process may produce ACE strings when its attempt to convert to the local encoding fails. See the section ``NAME DECODING'' for details.

(2) Local mapping

Apply any locale/language-specific mapping. This process is further divided into two subprocesses:

(2-1) Delimiter mapping

Map certain characters to the domain name delimiter (U+002E). This step is performed if actions given to mdn_encodename contains MDN_DELIMMAP or equals to MDN_ENCODE_APP.

(2-2) Local mapping based on TLD

Apply character mapping whose rule is determined by the TLD of the name.

This step is performed if actions given to mdn_encodename contains MDN_LOCALMAP or equals to MDN_ENCODE_APP.

(3) NAMEPREP

Perform name preparation (NAMEPREP), which is a standard process for name canonicalizaion of multilingual domain names. mdn_encodename performs mapping, normalization, prohibited character check and unassigned codepoint check in that order, as NAMEPREP mentions.

This process includes checking of characters which are not allowed in multilingual domain names. This step is performed if actions given to mdn_encodename contains MDN_NAMEPREP or equals to MDN_ENCODE_APP.

(4) Encoding conversion (UCS to ACE)

Convert the NAMEPREPed name to a special encoding designed for representing multilingual domain names.

The encoding is also known as ACE (ASCII Compatible Encoding) since a string in the encoding is just like a traditional ASCII domain name consisting of only letters, numbers and hyphens. This step is performed if actions given to mdn_encodename contains MDN_IDNCONV or equals to MDN_ENCODE_APP.

There are many configuration parameters for this process, such as the ACE or the local mapping rules. These parameters are read from the default mDNkit's configuration file, mdn.conf. See mdn.conf(5) for details.

Name decoding is a reverse process of the name encoding. It transforms the specified multilingual domain name in a special encoding suitable for name resolution to the normal name string in the application's current codeset. However, name encoding and name decoding are not symmetric. Name decoding doesn't perform local mapping. Both name encoding and decoding do NAMEPREP, but decoding does it to verfiy a name, while encoding does it to normalize a name.

The name decoding process consists of the following steps.

(1) Encoding conversion (ACE to UCS)

Convert the encoding of the given domain name from a special one designed for representing multilingual domain names (ACE) to Unicode/ISO10646 UTF-8. This step is performed if actions given to mdn_decodename contains MDN_IDNCONV or equals to MDN_DECODE_APP.

(2) NAMEPREP check

Perform name preparation (NAMEPREP) as also done in name encoding, and compare the result of NAMEPREP and the given UCS name. If the NAMEPREP is failed or the names are different, the given name is considered as invalid domain name. It contains a character which is never seen in a NAMEPREPed domain name. This step is performed if actions given to mdn_decodename contains MDN_NAMEPREP or equals to MDN_DECODE_APP.

(3) Encoding conversion (UCS to local codeset)

Convert the encoding of the name from UTF-8 to the application's local codeset.

The local codeset is determined by the application's locale. However, it is possible that the conversion to the application's local codeset fails because the name includes some characters which don't have mappings to the local codeset. In this case, libmdn tries instead to convert to ACE specified by the configuration file.

This step is performed if actions given to mdn_decodename contains MDN_LOCALCONV or equals to MDN_DECODE_APP. Note that libmdnlite doesn't support this step.

The configuration parameters for this process, are also read from the configuration file mdn.conf.

If the MDN_DISABLE environ variable is defined at run-time, the libraries disable mulitilingual domain name support, by default. In this case, mdn_encodename and mdn_decodename don't encode/decode an input name, but instead they simply ouput a copy of the input name as the result of encoding/decoding.

If your application should always enable mulitilingual domain name support regardless of definition of MDN_DISABLE, call

mdn_enable(1)

Most of the API functions return values of type mdn_result_t in order to indicate the status of the call.

The following is a complete list of the status codes. Note that some of them are never returned by the functions described in this manual.

Not an error. The call succeeded.

Specified information does not exist.

The encoding of the specified string is invalid.

There is a syntax error in the configuration file.

The specified name is not valid.

The specified DNS message is not valid.

The specified action contains invalid flags.

The specified Unicode code point value is not valid.

The specified buffer is too small to hold the result.

The specified key does not exist in the hash table.

Memory allocation using malloc failed.

The specified file could not be opened.

Some characters do not have the mapping to the target character set.

Context information is required.

The specified string contains some prohibited characters.

Generic error which is not covered by the above codes.

To get the address of a multilingual domain name in the application's local codeset, use mdn_encodename to convert the name to the format suitable for passing to resolver functions.

mdn_result_t r;
char ace_name[256];
struct hostent *hp;

... r = mdn_encodename(MDN_ENCODE_APP, name, ace_name, sizeof(ace_name)); if (r != mdn_success) { fprintf(stderr, "mdn_encodename failed: %s\n", mdn_result_tostring(r)); exit(1); }

hp = gethostbyname(ace_name); ...

To decode the multilingual domain name returned from a resolver function, use mdn_decodename.

mdn_result_t r;
char local_name[256];
struct hostent *hp;

... hp = gethostbyname(n); r = mdn_decodename(MDN_DECODE_APP, hp->h_name, local_name, sizeof(local_name)); if (r != mdn_success) { fprintf(stderr, "mdn_decodename failed: %s\n", mdn_result_tostring(r)); exit(1); } printf("name: %s\n", local_name); ...

mdn_encodename and mdn_decodename in mDNkit version 2.x prior to 2.4 provide the MDN_UNASCHECK process macro for unassigned codepoint check. However, beginning with mDNkit version 2.4, the MDN_NAMEPREP process macro also performs the unassigned check.

To perform NAMEPREP without the unassigned check on mDNkit 2.4 or later, specify the actions:

MDN_NAMEPREP^MDN_UNASCHECK

instead of MDN_NAMEPREP.

/etc/mdn.conf

mdn.conf(5)