csv2_txt - Man Page

Description of txt and raw resource records in the csv2 zone file


Due to the complexity of TXT and RAW records, this man page is dedicated to describing the csv2 format of this RR.

TXT and RAW rrs in MaraDNS' csv2 zone files can store any arbitrary binary data. Additionally, it is possible to arbitrarily divide up TXT records in to chunks (chunks, which RFC1035 call "character-string"s, are described below).


If a given TXT field or RAW record contains only ASCII data, creating a record is easy: Place the full data between single quotes, like this:

a.example.com. TXT 'This is some text' ~

It is also possible, to place almost any printable ASCII characters between quotes. The '~' (tilde) character is not allowed unless csv2_tilde_handling has a value of 0; the '|' (pipe), '#' (hash) and non-printable ASCII control characters are not allowed in TXT data if the ~ is used to separate records. If there are any bytes with a value of 0x80 or more, the data must be UTF-8 encoded Unicode.

The printable ASCII characters not allowed in quotes are the ' character, the '|' character, the '~' (tilde) character, and the '#' character. See BACKSLASH ESCAPE SEQUENCES below for information on adding these characters to TXT or RAW fields.


Note that the record does not have to be quoted. As long as the record only contains ASCII alphanumeric data, and/or the characters '-', '_', '+', '%', '!', '^', and '=', the data can be unquoted as follows:

c.example.com. TXT This_is_100%_unquoted_text_+symbols!

It is also possible to mix quoted and unquoted text, such as this:

d.example.com. TXT This' is a mix 'of_unquoted' and quoted 'text!

Which will have its data look like this:

This is a mix of_unquoted and quoted text!

When mixing quoted and unquoted data, it is important to have all whitespace inside quotes.


In order to accommodate storing non-UTF-8 high bit characters, the single quote character, non-printable ASCII control codes, the '|', '~', and '#' characters, and to permit multi-line TXT/RAW records (with comments allowed mid-record), the TXT/RAW RR allows backslashes. These backslashes only have significance outside of quoted text; if they are placed inside single quotes, they are not interpreted and result in a literal backslash being added to the resource record data.

The following characters can be backslashed:


When backslashed, the adds a literal quote to the resource record.


When any whitespace is backslashed (space, newline, cr, and tab), this indicates that the record has not ended, and that more data for this resource will follow. This also allows comments to be placed in TXT and RAW resource records. What happens is that the backslash indicates that any whitespace characters (space, tab, carriage return, and line feed) are to be ignored until the next non-whitespace character that is not a # (hash). If a # is seen, this indicates that we ignore any and all characters until the next carriage return or line feed, and continue to ignore everything until the next non-whitespace character. See the section on multi-line and commented records for examples.


When a number between 0 and 3 is backslashed, this indicates the beginning of a three-digit octal number.


When an x is backslashed, this indicates the beginning of a two-digit hexadecimal number.

Note that, with the exception of the single quote, the backslash character is not used to remove the meta-significance of a given character. In particular, unlike other environments, it is not possible to backslash spaces. Spaces can be represented either as ' ' in quotes, \x20, or as \040.

Here are some examples of backslashed data. In this example, we see backslash sequences being used to store non-UTF-8 hi-bit data:

e.example.com. TXT \x80\x81\x82\x83 ~

This same data can also be created as follows:

f.example.com. TXT \200\201\202\203 ~

Octal and hex information can be mixed:

g.example.com. TXT \200\x81\202\x83 ~

Literal single quotes can be placed in resource records:

h.example.com. TXT 'perl -e '\''print "A Perl of a TXT record!\n"'\' ~

The above example produces this record:

perl -e 'print "A Perl of a TXT record!\n"' ~

To render the '~' character, use the escape sequence \x7e (outside of quotes). For example:

h1.example.com. TXT 'http://ocf.berkeley.edu/'\x7e'set' ~

Produces this record:


To render the '|' character, use the escape sequence \x7c:

h2.example.com. TXT 'ls '\x7c' more' ~

Produces this record:

ls | more

To render the '#' character, use the escape sequence \x23:

h3.example.com. TXT 'Press '\x23' for customer service' ~

Produces this record:

Press # for customer service


By utilizing backslashes followed by comments, it is possible to have multi-line and commented TXT and RAW records. The following resource record will span more than one line on an 80-column display:

i.example.com. TXT 'Not only did the quick brown fox jump over the lazy dog, but the lazy dog jumped over the cat.' ~

Without affecting this resource record, the same data can be split over multiple lines:

j.example.com. TXT 'Not only did the quick brown fox jump '\
                   'over the lazy dog, but the lazy dog'\
                   ' jumped over the cat.' ~

Some points:

It is also possible to add comments after such a backslash as follows:

k.example.com. TXT 'Not only did the quick brown fox jump '\ # The fox
                   'over the lazy dog, but the lazy dog'\    # The dog
                   ' jumped over the cat.' ~                 # The cat

Note that, since the third comment is not preceded by a backslash, this indicates the end of the resource record.

There can also be multiple lines dedicated to comments (and, optionally, even blank lines) in the middle of TXT and RAW record data:

k2.example.com. TXT 'This is some data '\
# Here we have some comments followed by a blank line

# Now we have some more comments, 
# followed by the rest of the data
    'and this is the rest of the data' ~


TXT RRs may be divided up in to multiple "chunks" (RFC1035 calls these "character-string"s). A single chunk can be anywhere from zero to 255 bytes long. The default is to have one chunk, as follows:

o.example.com. TXT 'TXT record with only one chunk' ~

It is also possible to have a record with multiple chunks. Chunks are delimited by an unquoted ';' character:

p.example.com. TXT 'This is chunk one';'This is chunk two' ~


q.example.com. TXT 'This is chunk one';\   # Our first chunk
                    This_is_chunk_two;\    # Our second chunk
                   'This is chunk three' ~ # Our final chunk

Quoted ; characters simply add a ; to the record data.

If a single TXT chunk is longer than 255 bytes long, the csv2 parser will report an error in the zone file: Single TXT chunk too long

In order to resolve this, place unquoted ; characters in the record data so that each chunk is under 255 octets (bytes or characters) in length.

It is possible to have zero length chunks:

r.example.com. TXT 'chunk one';;'chunk three' ~ # Chunk two zero-length

In particular, is is possible to have zero length chunks at the beginning and end of a TXT record:

s.example.com. TXT ;'chunk two'; ~ # Chunks one and three zero-length

Do not place semicolons at the beginning nor end of TXT records unless you wish to have these zero-length chunks.

Chunk support only exists for TXT records. An unquoted ; character will cause a syntax error in a RAW record.


With the exception of no support for chunk delimiters, and the addition of a numeric record type before the record data, the format for RAW records is identical to text records. For example, if we wish to have a "Kitchen Sink" RR record, which has the 8-bit binary numbers "16", "1", and "2", followed by the ASCII string "Kitchen sink+ data", we can specify this in any of the following manners:

t1.example.com. RAW 40 \x10\x01\x02'Kitchen sink'\x2b' data' ~
t.example.com. RAW 40 \020\001\002Kitchen' sink+ data' ~
u.example.com. RAW 40 \x10\x01\x02Kitchen\x20sink+\x20data ~
v.example.com. RAW 40 \x10\001\x02\
                      'Kitchen sink+ data' ~
w.example.com. RAW 40 \x10\ # Meaning: 16
                      \x01\ # Coding: 1
                      \x02\ # Sub-coding: 2
                      'Kitchen sink+ data' ~ # Data: 'Kitchen sink+ data'


Sam Trenholme http://www.samiam.org/

Referenced By


January 2007 MaraDNS reference