Release Name: 1.16
==================== Quiet warnings on negative count from x.
	New encoding-setup: auto-detection of UTF-8 environments (currently via LANG only).
	New encoding-setup: force by calling reset_encode_decode_config('encoding') (at least before start of job).

	Sort with -R (taking into account the numbers inside file names).
	Allow track=... for -F.
	Auto-enable unicode processing of CMD line based on LANG (disable by MP3INFO_DECODE_ARGV_DEFAULT_RESET).
	Allow -e0, and default to -e15 if $LANG indicates utf-8 and there is no appropriate $^UNICODE flags.
		???  Should not we be more delicate with processing different flags of ${^UNICODE}???
	New option -U for defaulting to UNICODE unconditionally of LANG.
		(Needed since -C63 has lexical effects: io only work in the main file!)
	New bit 0x08 for -e to influence file i/o.
	Mention MP3TAG_NORMALIZE_FIELDS in the ENVIRONMENT section of the docs.
	When converting encoding "deep", do not crash on Regexp (needed for audio_rename).

	Misprint in docs.
	New option -U (as for mp3info2).
	New option -L (as -e for mp3info2, but without 'binary').

Release Name: 1.15
====================	Fix more unescaped-{ warnings (thanks to Christian Walde etc).

Release Name: 1.14
==================== Quiet warnings from 5.22.
	Start implementing handlers: instead of FRAM(langs)[opts], one can use, e.g., func_name(ID3v1,Cue)[arg1][arg2]
		   # ID3v2::_frame_select_by_descr is missing //s
	New configuration variable ampersand_joiner (default '; ').
	New method _auto_field_from() (abstracted from _auto_field()).
	Change logic of calling ->can() in _auto_field_from().
	_parse_rex_microinterpolate() would not update $ecount on seeing %%.
	When parsing with %=c etc: with %==c the match fails if there is no comment.
		(Checked the same way as for %{c:}.)
	Support some of %-escapes not being matched (e.g, due to alternatives in a REx).
		(0-length matches were ignored anyway [when join()ing].)
	parse_rex(), parse_rex_match() may return an extra result (if %{handler}s are present).
	Recognize ID3v2 frame names as [A-Z]{3}[A-Z\d] (was \w{4}).
		(Detection frame/vs/handler happens via PACKAGES; so if lang codes inf/cue appear, we may be in trouble.)
	Use the same code in parse(_rex)?_prepare.  (Now parse() allows the same %-constructs as parse_rex().)
	New method have_one_of_frames().
	New methods *_have() (for simplest fields: title, comment, track, artist, album, genre, year).
	New encoding-setup: auto-detection of UTF-8 environments (currently via LANG only).
	New encoding-setup: force by calling reset_encode_decode_config('encoding') (at least before start of job).
	Comprehensive docs.

Release Name: 1.13

Apply patches from Ian:
 Makefile.PL,data_pod.PL: Fix path of ID3v2-Data.pod. Fix #! path to perl.
 data_pod.PL,,,,,, Fix pod fix frame numbering documentation
 Many: preceeded/explicitely striking again ;-)

ID3v2:	add section BUGS with explanation of how we treat v2.4 frames.
	extract MP3 start bytes even if ID3v2 tag is present.
	recognize MP3 MIME type by start bytes	document mime_Pretype(), mime_type().
	New optional arguments to mime_Pretype(), mime_type().
	ID3v2 genre tag with value "Blues" or "(0)" could not be set
		(thanks to Eric Benson for analysis).

ID3v2_Data.pod: rename s/-/_/ - cygwin buggy, fails manifying
		- this was BS, the problem is elsewhere

mp3info2: state another time that COMM(eng)[lyrics] is a non-standard place
	  for lyrics...

Release Name: 1.12

Music_Normalize_Fields: new short names in RUS table.
  Eliminate one record which caused infinite recursion (vas i iva).

ID3v2: replace order of encode() and func_back() when writing frames
	(e.g., would break unicode genre with space in it).
       "very low-level" docs had get_frame() mixed up with get_frames()
		(thanks to Brendan Byrd)
       new_with_parent() inserts header into parent on failure.
       optional argument to _Data_to_MIME to return false on unknown. better docs for interpolate().
	misprints in example of update_tags().
	When truncating inside interpolation, and the value is a number
	  round it, not truncate.
	Move old meaning of %{aR} to %{a3} (truncation to 3 places).
	%{aR} is not rounded now; one may truncate as in %.5{aR} to restrict width.
	New method mime_Pretype(); new interpolation escape %{mP}.
	Optional argument to _Data_to_MIME() to return false on unknown.
	For (image) mime types we can handle ourselves, do it.
	default return for mime_type() is MP3 (in absense of support modules).
	Misprint in _Data_to_MIME() made it useless.
	New method aspect_ratio_inverted().
	New escape %{aI} for aspect_ratio_inverted().
	Allow for rounding of duration before formatting T[=>m,?H:,{mL}]. fix mismatched C<> in POD.

mp3info2: -F disk_n=DISKNUMBER supported, 
  new configuration variable frames_write_creates_dirs (default 0) to create
    intermediate directories baz/boz on -F "foo > baz/boz/bar".
  Use %{mP} instead of `Audio' in the -x output.
  Make -x=output of MP3-related data conditional on its presence (still wrong
    since MP3::Info can return such data for non-MP3 files too XXX)

	new option -r
	Make haveLyrics/haveAPIC symbols take less space
	Macroize pre/postSectionSkip, make it and \columnsep font-size dependent
	Slightly more docs in the generated code on grid/non-grid typesetting
	Slightly more docs

Release Name: 1.11
==================== new handler (a very minimal implementation). support for
  (problems: deleting frames is not noticed by ExifTool, so the field can be
     "revived" by update_tags(), so we disable id3 in ExifTool [XXX: must be
     made conditional on the list of  handlers]; genre-->None massaged to '')
             'NumColors' => 'Use BitDepth'???  
  rename aspect() to aspect_ratio() (undocumented)
  new function aspect_ratio3() rounds to 3 decimals in the fractional part
  map %{aR} to aspect_ratio3()  [instead of aspect()]
  new method bit_depth() (for images; exact semantic [8 or 24?!] not clear)  (undocumented)
  new interpolation %{bD} for bit_depth
  new method _duration() (for very few formats only???) (undocumented)
  caching MP3::Info data avoid extraneous calls
  Make duration-fetch methods try _duration() too.
  mpeg_layer_roman(), frames_padded_YN(), is_copyrighted_YN() now may return ''
    if MP3::Info fails.
  Allow for the rest of 1-char escapes NvLrqQSmsCpouMH to be conditioned and
  //o-ptimise RExes. could modify config('default_language').

mp3info2: make MP3-specific fields optional
  emit %wx%h for images and video (and append bit_depth if present)

Release Name: 1.10

Bump the version number to 1.10 to sync with mp3info2. allow %=n to match against 1 if %n is 01.
  new methods extension_is(), can_write(), can_write_or_die(), die_cant_write(),
  new methods (require Image::Size) width(), height(), img_type(), mime_type(),
    aspect(). (undocumented)
  new interpolation escapes %w %h %{iT} %{mT} %{aR} for these methods.
  new configuration variables: `is_writable' defaults to writable_by_extension
    (older versions behaved as if it is 1);  `writable_extensions' to
       mp3 mp2 id3 tag ogg mpg mpeg mp4 aiff flac ape ram mpc
  update_tags() checks for can_write().

cddb.tmp was omitted from distribution (==> test failure). the code to treat compressed fields was completely busted.
	  (Still not tested a lot...)
  Now always write UTF-16 as `UTF-16LE with BOM' to cowtow to M$'s bugs
	(must disable: triggers unsync...; set $ENV{MP3TAG_USE_UTF_16LE}
	 to TRUE to enable). could give `uninitialized' warning in presence of
	one-word person names in the tables. new handler.

mp3info2: -d was busted by the repeated-option logic.
  allow a repeated -u to force update.

typeset_audio_dir: if one of the arguments was 0, -e could fail.
  Decrease default min-gap before [year] from 0.75em to 0.65em (looks better).

Release Name: 1.00

Bump all the version numbers to 1.00 (except Normalize::Text::Music_Fields).

Move code to a ./lib subdirectory.  (Allows auto-install of transliterate_cp1251.) renamed to
  Now check environment variable MUSIC_FIELDS_PATH and/or ~/.music_fields/*.
  name_for_field_normalization method is used to choose "composer name". minor changes to the docs. new sub-module. allow access to Cue, and customization of the corresponding encoding.
  Allow conditional interpolation on user variables (check emptiness) %{U1:\n}.
  New configuration variable and method `name_for_field_normalization'.

typeset_audio_dir: Major simplification of user interface, better docs.
  Can generate short+long-type listings itself (two-pass so far...).
  More consistent ornaments in the generated files.
  Auto-generate the file *_titles.tex (used as the title of the listing).
  Move common code to file *_common.tex.
  New options -N, -P, -F.
  Document option -L.
  Directory list defaults to ".".
  Indicator of presence of APIC frames.
  Lyrics symbols explanation emitted correctly.
  Allow duplication of comment-like info and a part of the main record title
    (emphasizes this part of the title).
  Document known possible problems during typesetting.

  Since Normalize::Text::Music_Fields is always loadable now, one
    uses a different trigger sequence to load this module (environment
    variable and a directory ~/.music_fields are checked).
  With -e N, N in 1..3, do not translate the non-option ARGV
    (could conflict with interation with filesystems).
  New bit 0x4 of -e N used to influence translation of non-option ARGV.
  When @ARGV is re-encoded: translate back to 8-bit whatever is possible. when ordering frames in generated tags, sort APIC frames
  at the end (since the simplest reason for unsync-related bugs in other
  programs is misparse of APIC frames, and it is the *consequent* frames
  which are busted, no non-APIC frames would be affected.  (This is
  applicable in the case when other frames contain Latin-1 info, so
  the chance of unsync-sequences there is low...) 

Release Name: 0.9714
==================== misprint in POD.
  Add new variable MP3TAG_SKIP_WEAKEN for debugging (does not help with 5.005).
  Make documentation of encoding settings more clear.
  New configuration variable `extra_config_keys'.
  `n' and `I' in interpolation flags were broken.
  interpolate_with_flags(): take into account that ->track may be not a
    number, as "1/17". do not prepend "(C) " to empty TCOP fields.

mp3info2: -F "WHOLE > file" would not write if the given tags are not present.
  would skip -F "WHOLE ?< file" if corresponding tags were present (as opposed to fields).
  Long description for TDRC was missing.
  Long description was wrongly handled in __frames_as_printable().
  New configuration variable `empty-F-deletes'.
  The default now is to remove the frame if value -F is empty.
  Take into account `decode_encoding_files' when processing -F "FRAME < file"; was not updated during move to a subdirectory.

typeset_audio_dir: better docs.
  Typeset track number without second part, and allowing customization too. take into account that ->track may be not a number, as "1/17".

Release Name: 0.9713

mp3info2: -e was broken by support for duplicated-options.
  Better structuring of documentation.

Release Name: 0.9712
  _massage_genres(): 'num' is required if number is wished; new semantic with
  Make methods *_set() return TRUE.
  In dump of frames, put explanation of frames purposes.
  Improve the docs. add `func_back' to frame's field descriptor.
  add 're2b' to frame's field descriptor.
  When adding TCOM frame, massage it back to the v2.3 conventions; now one can
    use the same values for setting frame as returned when reading frame, even
    in presence of multiple genres.
  Likewise for TCOP frames (via `re2b').
  Repeated APIC frames were not expanded in get_frame_descriptors().
  Would not support "fake" MP3 objects.
  Allow setting of "filtered for human" fields (except TMED) via add_frame(),
    using both raw and human-readable forms of the corresponding argument.
  When splitting an argument for add_frame, split from the end for
    POPM - they have numbers at end, so it is not ambiguous.
  AENC and COMR frames had invalid format (trailing 0 omitted from URL).
  Dump of frames: put explanation of frames purposes, align stuff vertically.
  Optional string fields were treated with warnings...
  Improve the docs. reading v1 genre immediately after setting might return numeric value.

Test suite for changes (except encoding stuff and as_bin()).
Makefile.PL was clean()ing a shipped file.

Release Name: 0.9711
==================== last winamp genre was not supported;
  logic of checking fitting to ID3v1 was not accepting "from"ified tags;
  add more WinAmp genres (from table in MP3::Info).
  Avoid an elusive Perl (parser?) bug with
   can't use string ("") as an ARRAY ref while "strict refs" in use at MP3/Tag/ line 220.
  Use the leading part of track numbers of the form NUMBER/WHATEVER.
  logic of massaging genre was not accepting "from"ified tags.
  id3v23_unsync id3v23_unsync_size_w - better docs.
  Minor improvements to docs.
  New methods: "simplified" title_set(), artist_set(), album_set(),
      year_set(), comment_set(), track_set(), genre_set();
    _Data_to_MIME(), copy_id3v2_frames(),
    track1(), track2(), track0(), disk1(), disk2(), disk_alphanum().
  Prefer frame_select_by_descr_simpler() to frame_select_by_descr_simple()
    in select_id3v2_frame_by_descr(); so we can:
  remove special-casing of _Data in _interpolate(), and the first ("supported")
    frame of the given type can be interpolated via, e.g., "COMM", "APIC".
  remove setting $MP3::Info::try_harder: may loop infinitely...
  New interpolation: %{ID3v2-size}, %{ID3v2-pad}, %{ID3v2-stripped}, %{ID3v2-modified}
  Allow condition in conditional interpolation to be a 2-char (n1 etc) escape.
  Allow select_id3v2_frames() etc to write frames which require multiple
    arguments to add_frame(), either by adding extra arguments, or by joining
    arguments together with ";". [Will catch less errors, but is more flexible.]
  _massage_genres(): extra arg must now be 'num' if number is wished; new semantic
    with 'prefer_num'. misprint of getconfig1, and in printout of supported versions.
  Add some E<gt> to POD; better doc of APIC in frame_select().
  _Data_to_MIME() documented.
  Use "Unknown...  Error?" as string representation of unknown pictype in APIC.
  Do extra close/open before writing ID3v2, so that the size of ID3v1 tag
   is taken into account when calculating the "best" padding.
  New methods: frame_select_by_descr_simpler(), add_raw_frame() [undocumented],
    copy_frames(), add_frame_split().
  Supports optional fields in frames for read/write (missing
    fields not set by get_frame() etc, and should be omitted in call to add_frame).
  allow multiple values in "write" flavor of frame_select() (as in add_frame());
    moreover, since it calls add_frame_select(), allow joining fields into
    one with separator=";".  [Will catch less errors, but is more flexible.]
  For APIC/COMR frames, allow the numeric "type" argument to add_frame() to be
    specified both as a byte, and as a stringified number (no conflict possible).

audio_rename: move dependent module to
    Encode:: namespace.
  Change the default rename recipe to use disk number, and pad track number
    to width of the maximal track number (if known).

mp3info2: With -x, print "ID3v1: present" if applicable, and ID3v2 size(s).
  Documentation was using I(fimb) where I{fimbB} was needed.
  Move id3v2_frames_autofill() before -p and -x.
  Do not autofill unless -u is given, or other modifications are done to ID3v2.
  Uses foo_set() to process -a/-t/-l/-y/-c/-n/-g options, so no "sane value" check
    is done any more...  Also disallows "asyncroneous on-demand order of setting"
    of these fields (but it was hardly documentable anyway)...
  Document order of processing the options.
  New semantics "FRAME_spec < FILE", "FRAME_spec ?< FILE", "FRAME_spec > FILE"
    for -F option (including full-tag operations).
  Documentation on the usage strategy: Escalation of complexity.
  Options -d/-F/-C/-P may be repeated (old nonsense of gluing together via
    triplication is still supported, but deprecated).
  -x may be repeated (dumps frames). better wording in an error message.

Release Name: 0.9710

Some minimal changes to make test suite pass "better" on 5.005
(Music_Translate_Fields does not work; %F is not guarantied to get full name;
some proxying problems remains [leaks ==> files remain open];
t/update_tags.t works well under debugger, not otherwise...). mp3info2: support Language with parentheses, as in
 "Cover (back)" (when used as "Picture type").
 Document id3v2_frame_descriptors().
 New interpolation token: %{frames}, %{frames/SEPARATOR}.
 Allow %{ID3v2: ... } and ID3v1 as conditionals.
 Parsing of %{composer} etc inside false conditionals was wrong. is not confused by multipart genres (such as described by id3v2). extra $raw-ness for get_frame() (return array [=ordered hash], or
          array of values, or same with ignoring `decode_encoding_v2');
 get_frame() handles MCDI as a "simple frame";
 document get_frame_descr() and get_frame_descriptors();
 new argument $raw_ok for as_bin() to return original tag if unmodified;
 new method fix_frames_encoding() to make encoding standard-conformant;
 fix_frames_encoding() autocalled (controled by `decode_encoding_v2',
  `id3v2_fix_encoding_on_write' and `id3v2_fix_encoding_on_edit');
 `decode_encoding_v2' may be disabled per tag via
   frame TXXX[trusted_encoding0_v2] (in turn, this feature may be disabled
   by `ignore_trusted_encoding0_v2');
 frame TXXX[trusted_encoding0_v2] may be auto-added by fix_frames_encoding()
   (controlled by `id3v2_set_trusted_encoding0');
 special-field-readers now check their calling API (to allow future extensions);
 reading TCON normalizes genres, skips duplicates, and handles 0-separated
   genres (as mandated by v2.4) [thanks to neil verplank for the prototype
 reading/writing of v2.4 tags controlled by `prohibit_v24'/`write_v24';
 misprint in "Media (e.g. label side of CD)" of APIC corrected (backward
   compatibility value preserved);
 APIC frames handled by frame_select() (and derivatives); uses "Picture Type"
   instead of "Language", and auto-calculates "MIME type" for
 likewise for get_frame_descr(). simplified method get_config1() for accessing one-element configuration
 conversion milleseconds ==> seconds was wrong (!!!) when NVs are 80-bit;
 new configuration variables `id3v2_recalculate', `ignore_trusted_encoding0_v2',
   `id3v2_set_trusted_encoding0', `id3v2_fix_encoding_on_write',
   `id3v2_fix_encoding_on_edit', `prohibit_v24', `write_v24' with sane defaults.
 instead of MCDI frame, autofills TXXX[MCDI-fulltoc] (since MCDI is supposed
   to be the "short TOC", and I do not know how to reduce one to another,
   and know only how to easily read the "full TOC"; please advice if you know).

	[For best results, it is better to rename MCDI frame if it was
	 autofilled with v0.9709:
	   mp3info2 -u@P "mi/@{MCDI}/@{TXXX[MCDI-fulltoc]}///m//@{MCDI}" f.mp3

   Due to these changes, the mess of fixing broken encodings is more or
   less resolved now (before, it was easy to read tags with broken
   encoding, but not edit it).  Likewise, reading genre is much easier now
   (but setting multipart genre may still be some challenge).

[If there are some problems with reading v2.4 tags, I would like to hear about
 them; likewise, I beg for code donation for better writing of v2.4 tags.]

mp3info2: new option -N to switch off "smart-alec interface";
  -F allows human-oriented identifiers: composer,text_by,orchestra,conductor
  will only call autoinfo() if needed (useful for copying ID3 frames, so
    that they are not modified before copying);
  better logic in choosing when to emit "Performer:" line.
  beef up docs.

typeset_audio_dir: could try to print total time even without -T.

audio_rename: pacify some warnings; better docs.

New scripts in examples:
	inf_2fake_cddb cddb2cddb renamed to;
  a tiny shim module with old name added;
  subroutines renamed from *translate* to *normalize*;
  shim subroutines with old names added;
  minimal docs added;
  new subroutine read_composer_file() [cut out from load_composer()];
  new subroutine normalize_opus();
  better normalization of opus when followed by variants of `No.4';
  new subroutine _normalize_piece(): as normalize_piece(), but allows
    better control over (no) database lookup;
  better recognition of a key ("Piece; b#");
  support for mail-header-format databases;
  support for shorted opus-number ids;
  support for creation of mail-header-format databases;
  minimal testing suite (builtin);
  other minor improvements.

Composer files: converted to mail-header format; new files for Brahms,
Dvorak, Schnittke, Shostakovich added.  Moved to subdirectory
`Music_Normalize_Fields'; should be relative to

Release Name: 0.9709
==================== was reversing the "informativeness" test, thus one needed
to redo Music_Translate_Fields-L_van_Beethoven.comp; add missing sonata
names to the latter.

Improve docs for ENVIRONMENT.  New customization variable local_cfg_file,
new method parse_cfg() (reads ~/.mp3tagprc).  [not tested]

New interpolation directives iC and iI to access cddb_id, cdindex_id.
New methods cddb_id, cdindex_id, id3v2_frames_autofill.  New configuration
variable id3v2_frames_autofill.  Support for filling MCDI frame.

mp3info2: id3v2_frames_autofill() is called unless -d option is given.
(Most probably the result is not propagated back to file unless -u option
was given.) new method (and parse directive) cddb_id. new methods (and parse directives) cddb_id, cdindex_id (cddb_id
has the leading '0x' stripped). new methods cddb_id, cdindex_id (read the corresponding TXXX frames).

Splitting UCS-2 fields by short-\0 was wrong.

Workaround for ITunes bug was not complete: only size calculation was
conditional on a configuration variable, but not the actual unsyncronization... - new variant (untested).

typeset_audio_dir was ignoring -n option.

Extend matching of track number when parsing to 999 (not more to avoid
confusion with years); such tracks are, apparently, written mod 256 into
ID3v1 tag.

Release Name: 0.9708

Allow character devices for files (so /dev/null now "works"
  mp3info2 -D -a beethoven -t "piano in F#" /dev/null
will print the info deduced from per-composer databases).

total_secs_trunc() was not documented.

New methods total_hours(), leftover_secs_float(), leftover_secs_trunc(),
leftover_mins().  Interpolated as %H, %{SML}, %{SL}, %{mL}.  Interpolate
leftover_msec() as %{ML}.  New method format_time().  New interpolated
%{T[format1,format2...]}.  New class method new_fake().
'update_length' config variable was not writable.  New method
			XXXX Undocumented...

ID3v1: support new option encoded_v1_fits (default NOT as it was).

ID3v2: add_frame() was not documented to support encoding.  New configuration
variables default_language, default_descr_c.  New functions get_frame_desc(),

Music_Translate_Fields: normalizes title from unambiguous short form (order
significant, so "piano sonata" and "sonata for piano" still not synonimous).
Allow #dup_opus_rex, #no_opus_no, #prev_short in composer table.
New function normalize_file_lines().

Table of names for Beethoven expanded to include all the Op/WoO works;
dates are extracted from 3 sources.  Table for Gershwin.

New file examples/

mp3info2: processing of the first part of -E was completely wrong; only one
option could have been \\-interpolated.  New syntax -E +something to add to
the defaults.  New option -x (modelled after mp3info, plus frame descriptors).
If file name is empty, will work with a fake tag object.  Print (guessed)
composer and orchestra fields too.

audio_rename: increase the length of -s to 110 (len controlled by environment
variable).  Add .tag .mp3 to -e.
Add documentation.  New option -r.  Now is installed.

typeset_audio_dir: now explains the meaning of (L) and (S).

Release Name: 0.9707

%M was not recognized.

parsing for "%n" allows track numbers of the form 123/789 too; such track
numbers are considered as "not fit for ID3v1".

Document that set_id3v2_frame(), select_id3v2_frame() remove the old value(s).

New method interpolate_with_flags().  Make ...::Parse use this method.

New methods composer(), performer(), delete_tag(), _interpolate() in
MP3::Tag.  Allow customization via local configuration modules
MP3::Tag::User MP3::Tag::Site MP3::Tag::Vendor.

New interpolated escape %{I(FLAGS)VALUE}.

New configuration variables id3v23_unsync, composer, performer,
translate_composer, translate_performer.

New test t/interpolate.t.

Incompatible change: kill backslashitis in %-escapes:
    frame_select_by_descr() etc: no backslashitis whatsoever.
    interpolate(): no need to backwack all \{}[], only the innermost delimiter
      char, and only those backslashes which preseed the innermost delimiter.
    Likewise for parse(), _parse_rex_microinterpolate().
  Need to change '%{y||<%\{COMM(rus,EN,#1,)[foo]\}>}' to '%{y||<%{COMM(rus,EN,#1,)[foo]}>}'
	'%%02t_Title: `%012.12t\'; %{TLAN} %{TLAN01: have %\{TLAN01\}} %{!TLAN02:, do not have TLAN02}'
     to '%%02t_Title: `%012.12t\'; %{TLAN} %{TLAN01: have %{TLAN01}} %{!TLAN02:, do not have TLAN02}'
     to '%{!COMM(rus,EN,#1)[foo]:<%{COMM(rus,EN,#1,)[foo]}>}'
     to '%{COMM(rus,EN,#1)[foo]||<%{COMM(rus,EN,#1,)[foo]}>}'
     to '%{COMM(rus,EN,#1,)[foo]||<%{COMM(rus,EN,#1,)[foo]}>}'
     to '%{COMM01||<%{COMM(rus,EN,#1,)[foo]}>}'
     to	'%{TLEN||<%{COMM(rus,EN,#1,)[foo]}>}'
	$mp3->select_id3v2_frame_by_descr('TXXX[with[\]]', 'this is my TXXX[with[]]')
	 to $mp3->select_id3v2_frame_by_descr('TXXX[with[]]', 'this is my TXXX[with[]]')

    In mp3info2:
	      print $mp3->interpolate(exists $opt{p} ? $opt{p} : <<EOC);
      Title:   %-50t Track: %n
      %{TCOM:Composer: %\\{TCOM\\}
      }%{TEXT:Text: %\\{TEXT\\}
      Album:   %-50l Year:  %y
      Comment: %-50c Genre: %g
	      print $mp3->interpolate(exists $opt{p} ? $opt{p} : <<EOC);
      Title:   %-50t Track: %n
      %{TCOM:Composer: %{TCOM}
      }%{TEXT:Text: %{TEXT}
      Album:   %-50l Year:  %y
      Comment: %-50c Genre: %g

mp3info2: allow deletion of whole tags by -d; improve docs.

typeset_audio_dir: improve typesetting, allow indication whether lyrics
is present, of total duration.

Music_Translate_Fields: short_name works better for, e.g., Beethoven (allow
  the "from" part, as von/van/di/de); normalize key signature and the
  composition name (per composer file; set MUSIC_TRANSLATE_FIELDS_SKIP_WARNINGS
  to skip warnings for not detailed enough standard description).

Music_Translate_Fields-L_van_Beethoven.comp: new file; piano sonatas only
	perl -wle "for (1..32) {{open my $f, qq(> $_)} system qw(mp3info2 -a Beethoven -l), qq(piano sonata No. $_), $_}"

examples/empty.mp3 got a tag accidentally.
New file examples/empty_10sec.mp3
  perl -we "print qq(\0) x (8000*2*10)" | lame -b 8 -r -s 8 -m m - empty_10sec.mp3
	(For experiments with players which only show ID3 when playing.)

If INSTALLSCRIPT is present in Makefile.PL options, LIB will not disable
installation of scripts.

Release Name: 0.9706

The weaken()ing code was failing since weaken() does not return TRUE.  (Tested
for no obvious memory leaks.)

Allow for backwacked \[ \] \\ in descriptor of, e.g., COMM[descriptor].

New configuration options:
		   encode_encoding_v1 decode_encoding_v1 decode_encoding_v2
		   decode_encoding_inf decode_encoding_cddb_file
		   decode_encoding_files encode_encoding_files
	(en/decode_encoding_files for ParseData; applicable to non-binary i/o).
Initialized from corresponding environment variables

Document environment variables.

translate_title_collection etc configuration values were not settable
(neither were they documented...).

New methods of ID3v2: frame_select_by_descr(), frame_have_by_descr(),
frame_list_by_descr(), frame_select_by_descr_simple().

New methods have_id3v2_frame_by_descr(), select_id3v2_frame_by_descr(),
frame_translate(), frames_translate(), shorten_person() in MP3::Tag.
interpolate() significantly simplified; new interpolation %{shP{NAME}}.

New type of interpolation: frame ids connected by &.  Slightly more readable
documentation of interpolate() and parse_rex().

New configuration settings: person_frames, and translate_person
(supported only by select_id3v2_frame_by_descr()). much better detection of non-canonical names
(e.g., year removal), support for person names shortening.  New method
check_persons() for checking consistency of translation/shortening.
Piecewise translation for multiple authors in a field (separated by ', ',
'; ', or ' - ').  Will translate title_collection and other CDDB fields.

Music_Translate_Fields-rus.lst: better list of persons for bard repertoire
and audiobooks; allow "fixing" wrong years too; handle "translated" and
many more separators of multi-names; cursory support for cases (genitive);
support for disabling translation (#keep). transliterate cyrillic with unicode too.

New options -d, -e, -F for the script mp3info2; brush up the docs.

typeset_audio_dir: new option -c; new generated file *_backcover.tex; slightly
better conversion of ASCII to TeX; use 'textcomp' (another incompatible LaTeX
change...); new macros for skips about second-level headings; encapsulate
\addOnRight; ornaments in *_text.tex and *_cdbooklet.tex; commented out
support for title; begin merging together the code to generate different
documents; consistent columnseprule; better positioning of hrule's; template
to merge several lists into one document; document latex, dvips etc. support unicode cyrillic.

Release Name: 0.9705

XXXX Not documented yet; does not cover all aspects (filenames, external
     files, Inf, CDDB files).
	v2 ENCODE (to illegal encodings) is not implemented yet.

Default for MP3TAG_DECODE_UTF8 is 1 (used for encoding too).

Encoding argument for v2->add_frame() is (finally!) implemented.

Text field of TXXX, USER, USLT, OWNE were wrongly marked as not encoded.

Setting lang=XXX via frame_select() or id3v2_select_frame() would result
in lang=xxx.

frame_select() returns the count of found frames if it is used for removal
of frames.

New parameter value 'hash' for get_frame().

New method frames_list() (variant of frame_select()).

Sanity check for tag size parameter in ID3v2.

Sanity check for wide characters in the prepared raw tag.

New method ->size for ::File.

More examples for update_tags().

Music_Translate_Fields: only the parsing logic remains; the data is
moved into separate files (encoding supported) Music_Translate_Fields*.lst
(should live alongside the .pm file).

typeset_audio_dir: new option -e; unify the charset logic; add utf8.

Add example/empty.mp3 (some MP3 applications do not accept "really empty"
MP3s; this one is "almost empty").

Release Name: 0.9704

Document why using "encoding" and "TPE1" mindlessly is a bad idea.

Cleaning up read support of v2.4.

Better error message for version mismatch.

Do not warn "not supported" any more if the footer is found (how could it be
"supported" anyway, except for skipping, which is already done, [was wrong]?).

Extended header parsed for v2.4 too.	(untested)

Footer was treated wrong...  Apparently frame flags were extracted very wrong
as well... (untested)  Unknown frame flags are mapped to 'unknown_<letter>'.

Rename 'padding_size' field to 'buggy_padding_size' to avoid confusion with
ID3v2 padding (it is the padding after the tag [unrelated to the standard],
not the padding inside tag).

fits_tag(): return FALSE if data contains chars above 255.
update_tags(): better docs, new argument $force2.

mp3info2: in the standard report, "artist" was offset by one space.  New option
   -2.  README.shrink updated correspondingly.

typeset_audio_dir: more comments on duplex output.

eat_wav_mp3_header: new options -D for dry run, -M for finding .mp3 files.

Release Name: 0.9703
typeset_audio_dir: would not work without -B switch
	Add generation of cdbooklet with -B.
mp3info2: when reporting contents, print "Composer" and "Text" lines too.
	Without -I and -p, writes "Performer" line if TXXX{TPE1} is present.
	New option -r.
method interpolate() could return '' when ID3v2 tag was missing even if
	the tag would affect only a part of the output.
There was no way to inform the parser that it is OK if the parse data
	is coming from file, and the file does not exist.  Add 'F' flag...

Release Name: 0.9702
	new name;
	Fully document;
	Change the semantic of numbers in .top_heading files by 1;
	New option -n;
	Generate 1st level header for all "toplevel" dirs;
	TeXify ", ` and ';
	Add \hrule's for default toplevel headings.
	-@ was not working;
Correct rounding of total_secs_int().
Truncated version in total_secs_int_trunc().
Rename total_millisecs_int() to total_millisecs_int_fetch(),
	likewise for total_secs_fetch().
New method total_millisecs_int() (with caching).
Cache duration in ms.
New escape: %M (total millisecs, integer).
New confiration variable: update_length.
Will update TLEN field if needed (governed by `update_length').
Alternation inside escapes implemented.
New file README.shrink with recommendations on how to fix results of
the bug with extra unmarked 0-padding.
Remove trailing .pl from names of typeset_audio_dir, mp3info2.
Install typeset_audio_dir, mp3info2 unless one of options -n or LIB=
is given to Makefile.PL.

Release Name: 0.9701

-P option of mp3info2 would ignore trailing empty parse patterns
  (not very restrictive, since to succeed, pattern should have a set-field,
  or match-exactly field; workaround: append @={U0}, and do not set U0).

When updating ID3v2 tag without growing the file, the tag length would
not include padding; thus the following update would not use the padding ==>
tag size leak.

New configuration options: id3v2_minpadding, id3v2_sizemult, id3v2_shrink,
id3v2_mergepadding (the first two have new defaults: 128 and 512; was 2K and

Close file for write of ID3v2 tag as early as possible.

Release Name: 0.97

Better docs for add_frame(), set_id3v2_frame(), for configuration parameter
id3v2_frame_empty_ok, parse_minmatch.

frame_select() now returns undef if no frame is found; likewise for
select_id3v2_frame(); better behaviour with very low number of arguments.

When interpolating %{COMM[text]}, use lang='', not undef (otherwise
answer is returned as hash).

Interpolation would interpret second % in '%%02n' as fill character.

Flag 'b' of ParseData would disable whitespace stripping from the data to
match, but not from matches.

->get_config() would not work as class method.

Ask MP3::Info to 'try_harder'; needed with very long ID3v2 headers...

Decoding Unicode was not flexible enough: now configured by two environment
variables: MP3TAG_DECODE_UNICODE (default 1) to enable decoding; the
target of decoding is determined by MP3TAG_DECODE_UTF8: if 0 (default), decoded
values are byte-encoded UTF-8 (every Perl character contains a byte of
UTF-8 encoded string); otherwise it is a native Perl Unicode string.

Language=XXX is the *official* designation of unknown language.

Make dependency for ID3v2-Data.pod on Tag/

Warn in mp3info2 if -P is used without ParseData in autoinfo.  Document
example of attaching lyrics to audio file.

CDDB_File could generate warnings for files without DGENRE field.

type_mp3_dir: new macro \hourmark, exchange ' and '' for minutes/sec.

Release Name: 0.96

A bug fix by Ben Bennett for 2.4 support.

Having a genre of the form '(2345)Some new type' would emit a warning.

UTF encoded fields are decoded by default.  Set MP3TAG_ENCODING=0 in
environment to disable.

Add ID3v2-Data.pod to MANIFEST to avoid circular dependency (XXXX bug:
it is generated using the installed version of, not the
current version).

Release Name: 0.95

After removing several of repeated frames, get_frame(s) could get confused.
So could the code using get_frame(s).  Now better documented, and fixed.

New method ->is_modified() of ID3v2.

New method ->as_bin() of ID3v1 and ID3v2.

New methods ->is_id3v2_modified(), ->get_id3v2_frames(), ->select_id3v2_frame(),
->have_id3v2_frame(), ->get_id3v2_frame_ids() of MP3::Tag.

Some `next's needed to be replaced by `return' in mp3info2 and audio_rename.

mp3info2 would not decide that an edit is not needed if an ID3v2 frame was
modified directly.

Avoid re-parsing by ParseData more aggressively (10x speedup in one complicated
real-life example).

set_id3v2_frame() and update_tags() were assuming that get_tags() was already

File->write() could lead to disaster with -l.		!!!!

New options 'b', 'B', 'o', 'O', 'D' for ParseData.  With 'n', 'l' ParseData
does not ignore empty trailing lines any more.

New %-interpolations: %{ID3v1} for the tag as a whole, same for ID3v2.

Be more strict during interpolation: do not recognize WORD1 as a valid short
name of ID3v2 frame.

Two new example scripts:,

Mention example scripts in documentation of MP3::Tag.

Release Name: 0.94

The example of the "long" form of ID3v2-comment setting API has arguments swapped.

New methods ->frame_select(), ->frame_have() of ID3v2.

New methods ->pure_filetags() and ->update_tags() of MP3::Tag.

New config variables id3v2_missing_fatal (it was as if it was defaulting to
TRUE; now defaults to FALSE), parse_minmatch.

New test files in t/: v2_comments.t update_tags.t parser.t 

New escape %{COMM(rus,eng,xxx,#4)[short_description]} (chooses a comment with
"short_description" descriptor in the specified list of languages; #4 means
4th comment field, or COMM03; empty language means any language.)  Works
as a condition for conditional interpolation too, and as a target of parsing
and ParseData.  Similarly for other frames which have 'Language' and/or
'Description' fields.

Some POD formatting fixed.

Setting "extra" ID3v2 frames was not immediately visible (only after write
and re-read from the file).

COMM's short-description was not marked as 'encoded' (due to a misprint).


COMM's frame short-description now sits on the key "Description", not "short".
[If you want backward compatibility, check both keys.]

get_frame() considers frames the only content of which is a URL as a "simple
frame"; likewise for frame_select(), but it also considers frames with _Data
only as "simple frames".

Release Name: 0.93

More robust processing of "Recorded..." in CDDB_File.

Better error message when parse expression has some `=' omitted.

When parsing with nothing to set, would emit "Unsupported reg..." (regmatch
would return 1 elt if no capturing parens present).

When rewriting id3v2 tag, protect against $\ being set.

Move to t/mp3tags.t, add t/set_v2.t

Allow arbitrary escapes of the form %{ABCD} or %{ABCD03} in the method parse()
etc; here ABCD is a 4-letter word.

New config parameter id3v2_frame_empty_ok.

It is possible to set individual id3v2 frames via ParseData; use %{TIT1} or
some such.  Setting to an empty string deletes the frame if config
parameter id3v2_frame_empty_ok is true.

Uses TCOM or TPE3 to find "artist" ($mp3->artist) too.

New option -R (recurse into directories) for example scripts.

New example script type_mp3_dir to produce a TeX file with contents of
directories.  (Put files `.content_comment' into a directory to include comments
specific for the directory layout.)

Connect parts of comment by \n if \n is already present there.

`Recorded: 1980,1985,1992' was not recognized.

Allow date only at start/end of the comment.

Release Name: 0.92

Include a fix by Marco Moreno of Cwd::abspath() not working on files (on
some architectures).

Release Name: 0.91

CDDB_File was removing whitespace on the boundary of continuation fields.

get_user() calls ParseData->parse() to populate the userdata

if $ENV{MP3TAG_TEST_WEAKEN}, will not have a workaround against broken weakrefs
	(currently tests fail!)

New %-escapes v L r q Q S m s C p o u (require MP3::Info module).

New methods mpeg_version() mpeg_layer() mpeg_layer_roman() is_stereo()
is_vbr() bitrate_kbps() frequency_Hz() frequency_kHz() size_bytes()
total_secs() total_secs_int() total_mins() leftover_secs()
leftover_msec() time_mm_ss() is_copyrighted() is_copyrighted_YN()
frames_padded() frames_padded_YN() channel_mode_int() frames()
frame_len() vbr_scale() (require MP3::Info module).

The year() method of ID3v2 takes/returns timestamps in human-readable form;
parse() recognizes the same formats for %y if option year_is_timestamp is TRUE

New option 'intact' to get_frame() of ID3v2.

Support YEAR, ID3Y, ID3G in EXTD, EXTT* fields of CDDB_File.  Support also
an alternative syntax "Recorded"/"Recorded on"/"Recorded in"/ with the format
of the date recognized by ID3v2::year(), or just a date field without a prefix.
The declarations of the former form are stripped from the returned comment.

New fields artist_collection, title_track, comment_collection, comment_track
in CDDB_File and corresponding accessor methods in MP3::Tag, and corresponding
escapes aC, tT, cC, cT for interpolate().

New configuration variables 'year_is_timestamp', 'comment_remove_date',
'translate_*' (for different values of *).

New handler module LastResort; currently it uses artist_collection() as comment
if comment is not otherwise defined.

New example scripts audio_rename and mod/

The example script mp3info2 will use (if found)
to postprocess the resulting tag fields via translate_*() methods.

Do not load Compress::Zlib unless when processing compressed data.

Release Name: 0.9

Revision history for Perl modules MP3::Tag, MP3::Tag::ID3v1, MP3::Tag::ID3v2, MP3::Tag::File,
             and for Perl program tk-tag


* Bugfix for
  - Added support for reading ID3v2.2 and ID3v2.4 tags (alpha stage!!) and converting them to ID3v2.3
  - Bugfix for unsynchronization (thanks to Ian Beckwith)
  - Bugfix for number-handling in frames
  - Bugfix for frame compression
  - Set unsynchronization flag only, if unsynchronization was done and changed anything

* Bugfix for
  - genres didn't return the id if the genre name was passed as parameter, but returned
    the same name

* Makefile.PL:
  - overriding the manifypods() function in MakeMaker to add ID3v2-Data.pod to the MAN3PODS hash
    (thanks to Dagfinn Ilmari Mannsaker)

* Added copyright text, using Perl Artistic License

Thanks to Ilya Zakharevich for the following changes:

* Three new modules MP3::Tag::Inf, MP3::Tag::ParseData, and
  MP3::Tag::CDDB_File added

  - autoinfo() method returns the info for all ID3v1 tags;
  - autoinfo() method may return the info which tag is obtained from
    which source;
  - Methods title(), author() etc (one per each ID3v1 tag) exist and
    have uniform interface through all the subpackages;
  - method song() is renamed to title(); method read_filename() is
    renamed to parse_filename(); backward compatible name still
  - config() subroutine can now change all the arbitrary decisions;
  - new methods interpolate() for tag interpolation into strings and
    parse(), parse_rex() for an inverse operation;
  - new subpackage `__hasparent' to treat cyclic object dependencies;
  - per-object configuration;
  - new configuration options extension, parse_data, parse_split,
    parse_filename_ignore_case, parse_filename_merge_dots, parse_join;
  - a way to store and query user-supplied data inside the object (may
    be used as scratch space when parsing);
  - new methods get_config(), get_user(), set_user(), parse_prepare(),
    parse_rex_prepare(), parse_rex_match(), filename(), abs_filename(),
    filename_nodir(), filename_noextension(), filename_nodir_noextension(),
    abs_filename_noextension(), dirname(), filename_extension(),
    filename_extension_nodot(), dir_component().

  - new ID3v1 method fits_tags() to check whether the info can be
    placed into an ID3v1 tag
  - If no track given use automatically ID3v1.0 tag with longer comment
  - open() file if needed; do not close() close during new().

  - allow multiple frames to be returned by get_frame() in ID3v2;
  - new method get_frames() with better order of returned values (comparing
    to get_frame);
  - get_frame_ids() in ID3v2 improved to take new argument 'truename' to
    simplify looping over repeated frames;
  - simpler treatement of sync; fix uncorrect calculation of padding length;
  - overwrite the tail of the previously present tag with 0s;
  - a couple of new v2.4 field names added.

  - more robust parsing of filenames; year can be read from filename if
    appended (in parentheses) to author or title; filename can be
    F<.wav> as well.
  - support file names starting with track numbers as in 03_This_is_the_title.mp3
  - support track_title.ext and title_track.ext formats with one-word title too.
  - new method filename();
  - use the 'extension' configuration variable to strip extension.

* misprints in the docs corrected and some general harmonization of function and
  field names in the different modules

* new example script mp3info2 (which provides most of functionality of the
  "standard" mp3info utility, and much more).

Release Name: 0.40

* Updated documentation for MP3::Tag, MP3::Tag::ID3v1, MP3::Tag::ID3v2 and MP3::Tag::ID3v2-Data 

* Renamed some functions. The old names will still work, but try to use the new names.
  The following names were changed: 

  - MP3::Tag::getTags() to MP3::Tag::get_tags() 
  - MP3::Tag::newTag() to MP3::Tag::new_tag() 
  - MP3::Tag::ID3v1::removeTag() to MP3::Tag::ID3v1::remove_tag() 
  - MP3::Tag::ID3v1::writeTag() to MP3::Tag::ID3v1::write_tag() 
  - MP3::Tag::ID2v2::getFrameIDs() to MP3::Tag::ID3v2::get_frame_ids() 
  - MP3::Tag::ID2v2::getFrame() to MP3::Tag::ID3v2::get_frame() 
* Bugfix for
  - getFrame() returned "undef" as a string instead of simple undef
  - artist() produced an error when TPE1 is missing in TAG
* Bugfix for
  - DESTROY() checked only if $mp3->obj{filename} exists and not if it is defined before trying to 
    close it  

* Bugfix for
  - genres() expected an id3v1-object as first parameter and a genre
    only as second parameter. Now the object can be omitted as in a call like
    MP3::Tag::ID3v1->genres($genre) used by and

* bugfix for
  - Filenames may contain surrounding whitespaces now

Release Name: 0.30
  - autoinfo() function added. This returns artist/songtitle/track/album.
    It tries to find this information in an ID3v1 or ID3v2 tag or tries 
    to extract it from the filename. The order in which this happens
    can be configured with the new config() function.
  - four new functions: artist(), song(), track() and album(). These
    are included for compability with the ID3v1 and filename module.
* /
  - All file access routines are collected in an extra modul now.
    This prevents circular references between and the
    modules. These circular references prevented Perl from calling
    the destructor for mp3-objects.
* tk-tag
  - Loading/Saving of binary data in frame-fields is supported now
  - tk-tag uses now the new what_data functionally of ID3v2 to offer 
    BrowseEntrys for TCON and TFLT and also for parts of APIC and COMR
  - Set Filename uses now the actual contents of the Tags not the old 
    saved one
  - Set Filename from ID3v2-Tag works now at least with Artist (%a),
    Album(%l) and Song (%s)
* ID3v2::what_data returns now also information about possible restricted
  input for some frame fields (APIC; TCON; COMR; TFLT are supported yet).

Release Name: 0.25

* Bug-fix for MP3::Tag
  If you created several mp3-objects for different files, the filehandles
  for each file were not used correctly. Thanks to hakimian for reporting
  this bug.
* Bug-fix for ID3v2::remove_tag()
  It was tried twice to rename one temp-file after removing the tag.
  Thanks to Brent Sarten <> for reporting this.
* Bug-fix for ID3v2::add_tag()
  When adding a second (or third, ...) frame of a kind, a wrong header
  could be written for this frame
* Bug-fix for tk-tag 
  When opening a new file, the contents of the ID3v2 tag of the last file
  would be copied to the ID3v2 tag of the new file for all frames which
  existed in the last file, but not in the new one.

Release Name: 0.21

* tk-tag has lots of new functions (see tk-tag README)
* getFrameIDs returns now always a hash ref of all existing frames, not only
  when called the first time
* Some minor problems in documentation are corrected.
* ID3v2::add_frame returns the name of the newly created frame, which can
  differ from the short name when already such a frame exists.

Release Name: 0.20 (beta)

--due to problems when run at windows:
  * Added a second seek to ID3v1::write_tag, as windows writes at a wrong 
    position otherwise
  * Setting Filehandle to binmode after opening a mp3 file 
  * ID3v2: write_tag creates a temp file (if neccessary) now in the same
    directory where the original mp3 files is located and not in /tmp

* Added, a graphical interface for MP3::Tag. is a alpha
* Added a new manpage MP3::Tag::ID3v2-Data which contains information about
  the ID3v2 frames and the data returned by MP3::Tag::Id3v2::getFrame() 
* Frames RVRB ("Reverb"), COMR ("Commercial frame"), AENC ("Audio encryption"),
  GRID ("Group identification registration"), RBUF ("Recommended buffer size")
  and SYTC ("Synchronized lyric/text") are now supported	
* Added some test to for creating new tags
* ID3v2::getFrameIDs returns now a hash reference, which contains the found
  frames. The keys are the 4 byte codes of the frames, which are needed for
  getFrame . The according values are the english (long) names of the frames.
* ID3v2::write_tag - Updating tagsize after writing tag 
* ID3v1::all() returns in array context all fields, otherwise only the song
* MP3::ID3v1::write_tag didn't returned an error if a file couldn't be opened
  for writing. Now it does.
* Renamed MP3::TAG to MP3::Tag following a suggestion of ANDK from CPAN
* Makefile.PL : Added that Compress::Zlib and File::Basename is needed for 
  installation of MP3::Tag

Release Name: 0.1 (beta)
* Added documentation to the modules

* Writing/removing of ID3v2.3 tags is supported now 
* Adding, changing, removing frames of ID3v2.3 is supported

* Changed directory structure
* Added file for proper install of modules

Release Name: 0.2-alpha
* ID3v2.3 compressed frames are supported now 
* changed directory structure, support librarys for MP3::Tag are now in a
* calls xview to show pictures, which were found in ID3v2 tags
  (sorry, not configurable at the moment, but easy to change in

Release Name: 0.1-alpha
This is the first alpha version. It contains perl modules to
read ID3v1/ID3v2 tags, but they are still lacking a lot of

* Reading / Writing ID3v1 works
* Reading of most frames of ID3v2.3 works

Included is a demo program, and a program to change ID3v1
tags and to set automatically the filename of a mp3 file: See
README.txt for details. More documentation is still lacking. Sorry.

MP3::Tag can be found at