draft-ietf-idn-lace-01.txt

bind-3.2.

section.

2.4.1 Compressing a string

The input string is in the UTF-16 encoding (big-endian UTF-16 with no
byte order mark).

Design note: No checking is done on the input to this algorithm. It is
assumed that all checking for valid ISO/IEC 10646 characters has already
been done by a previous step in the conversion process.

1) If the length (measured in octets) of the input is not even, or is
less than 2, stop with an error.

2) Set the input pointer, called IP, to the first octet of the input
string.

3) Set the variable called HIGH to the octet at IP.

4) Determine the number of contiguous pairs at or after IP that have
HIGH as the first octet; call this COUNT.

5) Put into an output buffer the single octet for COUNT followed by the
single octet for HIGH, followed by all those low octets. Move IP to the
end of those pairs; that is, set IP to IP+(2*COUNT).

6) If IP is not at the end of the input string, go to step 3.

7) If the length of the output buffer is less than or equal to the
length of the input buffer (in octets, not in characters), emit the
output buffer. Otherwise, output the octet 0xFF followed by the input
buffer. Note that there can only be one possible representation for a
name part, so that outputting the wrong name part is a serious security
error. Decompression schemes MUST accept only the valid form and MUST
NOT accept invalid forms.

2.4.2 Decompressing a string

1. Set the input pointer, called IP, to the first octet of the input
string. If there is no first octet, stop with an error.

2. If the octet at IP is 0xFF, set IP to IP+1, copy the rest of the
input buffer to the output buffer, and go to step 9.

3. Get the octet at IP, call it COUNT. If COUNT equals zero or is
greater than 36, stop with an error. Set IP to IP+1. If IP is now at the
end of the input string, stop with an error.

4. Get the octet at IP, call it HIGH. Set IP to IP+1.

5. If IP is now at the end of the input string, stop with an error. Get
the octet at IP, call it LOW. Set IP to IP+1.

6. Output HIGH, then LOW, to the output buffer.

7. Decrement COUNT. If COUNT is greater than 0, go to step 5.

8. If IP is not at the end of the input buffer, go to step 3.

9. If the length of the output buffer is odd, stop with an error.
Compress the output buffer into a separate comparison buffer following
the steps for compression above. If the contents of the comparison
buffer does not equal the input to the compression step, stop with an
error. Otherwise, send out the output buffer and stop.

2.4.3 Compression examples

The five input characters <U+30E6 U+30CB U+30B3 U+30FC U+30C9> are
represented in big-endian UTF-16 as the ten octets <30 E6 30 CB 30 B3 30
FC 30 C9>. All the code units are in the same row (03). The output
buffer has seven octets <05 30 E6 CB B3 FC C9>, which is shorter than
the input string. Thus the output is <05 30 E6 CB B3 FC C9>.

The four input characters <U+012F U+0111 U+0149 U+00E5> are represented
in big-endian UTF-16 as the eight octets <01 2F 01 11 01 49 00 E5>. The
output buffer has eight octets <03 01 2F 11 49 01 00 E5>, which is the
same length as the input string. Thus, the output is <03 01 2F 11 49 01
00 E5>.

The three input characters <U+012F U+00E0 U+014B> are represented in
big-endian UTF-16 as the six octets <01 2F 00 E0  01 4B>. The output
buffer is nine octets <01 01 2F 01 00 E0 01 01 4B>, which is longer than
the input buffer. Thus, the output is <FF 01 2F 00 E0 01 4B>.

2.5 Base32

In order to encode non-ASCII characters in DNS-compatible host name parts,
they must be converted into legal characters. This is done with Base32
encoding, described here.

Table 1 shows the mapping between input bits and output characters in
Base32. Design note: the digits used in Base32 are "2" through "7"
instead of "0" through "6" in order to avoid digits "0" and "1". This
helps reduce errors for users who are entering a Base32 stream and may
misinterpret a "0" for an "O" or a "1" for an "l".

                    Table 1: Base32 conversion
             bits   char  hex         bits   char  hex
             00000   a    0x61        10000   q    0x71
             00001   b    0x62        10001   r    0x72
             00010   c    0x63        10010   s    0x73
             00011   d    0x64        10011   t    0x74
             00100   e    0x65        10100   u    0x75
             00101   f    0x66        10101   v    0x76
             00110   g    0x67        10110   w    0x77
             00111   h    0x68        10111   x    0x78
             01000   i    0x69        11000   y    0x79
             01001   j    0x6a        11001   z    0x7a
             01010   k    0x6b        11010   2    0x32
             01011   l    0x6c        11011   3    0x33
             01100   m    0x6d        11100   4    0x34
             01101   n    0x6e        11101   5    0x35
             01110   o    0x6f        11110   6    0x36
             01111   p    0x70        11111   7    0x37

2.5.1 Encoding octets as Base32

The input is a stream of octets. However, the octets are then treated
as a stream of bits.

Design note: The assumption that the input is a stream of octets
(instead of a stream of bits) was made so that no padding was needed.
If you are reusing this algorithm for a stream of bits, you must add a
padding mechanism in order to differentiate different lengths of input.

1) Set the read pointer to the beginning of the input bit stream.

2) Look at the five bits after the read pointer. If there are not five
bits, go to step 5.

3) Look up the value of the set of five bits in the bits column of
Table 1, and output the character from the char column (whose hex value
is in the hex column).

4) Move the read pointer five bits forward. If the read pointer is at
the end of the input bit stream (that is, there are no more bits in the
input), stop. Otherwise, go to step 2.

5) Pad the bits seen until there are five bits.

6) Look up the value of the set of five bits in the bits column of
Table 1, and output the character from the char column (whose hex value
is in the hex column).

2.5.2 Decoding Base32 as octets

The input is octets in network byte order. The input octets MUST be
values from the second column in Table 1.

1) Count the number of octets in the input and divide it by 8; call the
remainder INPUTCHECK. If INPUTCHECK is 1 or 3 or 6, stop with an error.

2) Set the read pointer to the beginning of the input octet stream.

3) Look up the character value of the octet in the char column (or hex
value in hex column) of Table 1, and add the five bits from the bits
column to the output buffer.

4) Move the read pointer one octet forward. If the read pointer is not
at the end of the input octet stream (that is, there are more octets in
the input), go to step 3.

5) Count the number of bits that are in the output buffer and divide it
by 8; call the remainder PADDING. If the PADDING number of bits at the
end of the output buffer are not all zero, stop with an error.
Otherwise, emit the output buffer and stop.

2.5.3 Base32 example

Assume you want to encode the value 0x3a270f93. The bit string is:

3   a    2   7    0   f    9   3
00111010 00100111 00001111 10010011

Broken into chunks of five bits, this is:

00111 01000 10011 10000 11111 00100 11

Padding is added to make the last chunk five bits:

00111 01000 10011 10000 11111 00100 11000

The output of encoding is:

00111 01000 10011 10000 11111 00100 11000
  h     i     t     q     7     e     y
or "hitq7ey".


3. Security Considerations

Much of the security of the Internet relies on the DNS. Thus, any
change to the characteristics of the DNS can change the security of
much of the Internet. Thus, LACE makes no changes to the DNS
itself.

Host names are used by users to connect to Internet servers. The
security of the Internet would be compromised if a user entering a
single internationalized name could be connected to different servers
based on different interpretations of the internationalized host
name.

LACE is designed so that every internationalized host name part
can be represented as one and only one DNS-compatible string. If there
is any way to follow the steps in this document and get two or more
different results, it is a severe and fatal error in the protocol.


4. References

[IDNComp] Paul Hoffman, "Comparison of Internationalized Domain Name Proposals",
draft-ietf-idn-compare.

[IDNReq] James Seng, "Requirements of Internationalized Domain Names",
draft-ietf-idn-requirement.

[ISO10646] ISO/IEC 10646-1:1993. International Standard -- Information
technology -- Universal Multiple-Octet Coded Character Set (UCS) --
Part 1: Architecture and Basic Multilingual Plane.  Five amendments and
a technical corrigendum have been published up to now. UTF-16 is
described in Annex Q, published as Amendment 1. 17 other amendments are
currently at various stages of standardization. [[[ THIS REFERENCE
NEEDS TO BE UPDATED AFTER DETERMINING ACCEPTABLE WORDING ]]]

[RFC2119] Scott Bradner, "Key words for use in RFCs to Indicate
Requirement Levels", March 1997, RFC 2119.

[RFC2781] Paul Hoffman and Francois Yergeau, "UTF-16, an encoding of ISO
10646", February 2000, RFC 2781.

[STD13] Paul Mockapetris, "Domain names - implementation and
specification", November 1987, STD 13 (RFC 1035).

[Unicode3] The Unicode Consortium, "The Unicode Standard -- Version
3.0", ISBN 0-201-61633-5. Described at
<http://www.unicode.org/unicode/standard/versions/Unicode3.0.html>.


A. Acknowledgements

Rick Wesson pointed out some error conditions that need to be
tested for. Scott Hollenbeck pointed out some errors in the
compression.

Base32 is quite obviously inspired by the tried-and-true Base64
Content-Transfer-Encoding from MIME.


B. Sample code

The following is sample Javascript code for the LACE algorithm.
This code is believed to be correct, but there may be errors in
it. The code is provided as-is and comes with no warranty of
fitness, correctness, blah blah blah.

/**
 * Converts to LACE compression format (without Base32) from
 *   UTF-16BE array
 * @parameter iArray Array of bytes in UTF16-BE
 * @parameter iCount Number of elements. Must be 0..63
 * @parameter oArray Array for output of LACE bytes.
 *   Must be at least 100 octets long to provide internal working space
 * @return Length of output array used
 * @parameter parseResult output error value if any
 * @author Mark Davis
 */

function toLACE(iArray, iCount, oArray, parseResult) {
//debugger;
  if (iCount < 1 || iCount > 62) 讃
    parseResult.set("Lace: count out of range", iCount);
    return;
  }
  if ((iCount % 2) == 1) 讃
    parseResult.set("Lace: odd length, can't be UTF-16", iCount);
    return;
  }
  var op = 0;