summaryrefslogtreecommitdiff
path: root/Master/tlpkg/tlperl/lib/Locale/Maketext.pod
blob: 15533e4e870b2a7a822ed2582be913fd980db507 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326

# Time-stamp: "2004-01-11 18:35:34 AST"

=head1 NAME

Locale::Maketext - framework for localization

=head1 SYNOPSIS

  package MyProgram;
  use strict;
  use MyProgram::L10N;
   # ...which inherits from Locale::Maketext
  my $lh = MyProgram::L10N->get_handle() || die "What language?";
  ...
  # And then any messages your program emits, like:
  warn $lh->maketext( "Can't open file [_1]: [_2]\n", $f, $! );
  ...

=head1 DESCRIPTION

It is a common feature of applications (whether run directly,
or via the Web) for them to be "localized" -- i.e., for them
to a present an English interface to an English-speaker, a German
interface to a German-speaker, and so on for all languages it's
programmed with.  Locale::Maketext
is a framework for software localization; it provides you with the
tools for organizing and accessing the bits of text and text-processing
code that you need for producing localized applications.

In order to make sense of Maketext and how all its
components fit together, you should probably
go read L<Locale::Maketext::TPJ13|Locale::Maketext::TPJ13>, and
I<then> read the following documentation.

You may also want to read over the source for C<File::Findgrep>
and its constituent modules -- they are a complete (if small)
example application that uses Maketext.

=head1 QUICK OVERVIEW

The basic design of Locale::Maketext is object-oriented, and
Locale::Maketext is an abstract base class, from which you
derive a "project class".
The project class (with a name like "TkBocciBall::Localize",
which you then use in your module) is in turn the base class
for all the "language classes" for your project
(with names "TkBocciBall::Localize::it", 
"TkBocciBall::Localize::en",
"TkBocciBall::Localize::fr", etc.).

A language class is
a class containing a lexicon of phrases as class data,
and possibly also some methods that are of use in interpreting
phrases in the lexicon, or otherwise dealing with text in that
language.

An object belonging to a language class is called a "language
handle"; it's typically a flyweight object.

The normal course of action is to call:

  use TkBocciBall::Localize;  # the localization project class
  $lh = TkBocciBall::Localize->get_handle();
   # Depending on the user's locale, etc., this will
   # make a language handle from among the classes available,
   # and any defaults that you declare.
  die "Couldn't make a language handle??" unless $lh;

From then on, you use the C<maketext> function to access
entries in whatever lexicon(s) belong to the language handle
you got.  So, this:

  print $lh->maketext("You won!"), "\n";

...emits the right text for this language.  If the object
in C<$lh> belongs to class "TkBocciBall::Localize::fr" and
%TkBocciBall::Localize::fr::Lexicon contains C<("You won!"
=E<gt> "Tu as gagnE<eacute>!")>, then the above
code happily tells the user "Tu as gagnE<eacute>!".

=head1 METHODS

Locale::Maketext offers a variety of methods, which fall
into three categories:

=over

=item *

Methods to do with constructing language handles.

=item *

C<maketext> and other methods to do with accessing %Lexicon data
for a given language handle.

=item *

Methods that you may find it handy to use, from routines of
yours that you put in %Lexicon entries.

=back

These are covered in the following section.

=head2 Construction Methods

These are to do with constructing a language handle:

=over

=item *

$lh = YourProjClass->get_handle( ...langtags... ) || die "lg-handle?";

This tries loading classes based on the language-tags you give (like
C<("en-US", "sk", "kon", "es-MX", "ja", "i-klingon")>, and for the first class
that succeeds, returns YourProjClass::I<language>->new().

If it runs thru the entire given list of language-tags, and finds no classes
for those exact terms, it then tries "superordinate" language classes.
So if no "en-US" class (i.e., YourProjClass::en_us)
was found, nor classes for anything else in that list, we then try
its superordinate, "en" (i.e., YourProjClass::en), and so on thru 
the other language-tags in the given list: "es".
(The other language-tags in our example list: 
happen to have no superordinates.)

If none of those language-tags leads to loadable classes, we then
try classes derived from YourProjClass->fallback_languages() and
then if nothing comes of that, we use classes named by
YourProjClass->fallback_language_classes().  Then in the (probably
quite unlikely) event that that fails, we just return undef.

=item *

$lh = YourProjClass->get_handleB<()> || die "lg-handle?";

When C<get_handle> is called with an empty parameter list, magic happens:

If C<get_handle> senses that it's running in program that was
invoked as a CGI, then it tries to get language-tags out of the
environment variable "HTTP_ACCEPT_LANGUAGE", and it pretends that
those were the languages passed as parameters to C<get_handle>.

Otherwise (i.e., if not a CGI), this tries various OS-specific ways
to get the language-tags for the current locale/language, and then
pretends that those were the value(s) passed to C<get_handle>.

Currently this OS-specific stuff consists of looking in the environment
variables "LANG" and "LANGUAGE"; and on MSWin machines (where those
variables are typically unused), this also tries using
the module Win32::Locale to get a language-tag for whatever language/locale
is currently selected in the "Regional Settings" (or "International"?)
Control Panel.  I welcome further
suggestions for making this do the Right Thing under other operating
systems that support localization.

If you're using localization in an application that keeps a configuration
file, you might consider something like this in your project class:

  sub get_handle_via_config {
    my $class = $_[0];
    my $chosen_language = $Config_settings{'language'};
    my $lh;
    if($chosen_language) {
      $lh = $class->get_handle($chosen_language)
       || die "No language handle for \"$chosen_language\" or the like";
    } else {
      # Config file missing, maybe?
      $lh = $class->get_handle()
       || die "Can't get a language handle";
    }
    return $lh;
  }

=item *

$lh = YourProjClass::langname->new();

This constructs a language handle.  You usually B<don't> call this
directly, but instead let C<get_handle> find a language class to C<use>
and to then call ->new on.

=item *

$lh->init();

This is called by ->new to initialize newly-constructed language handles.
If you define an init method in your class, remember that it's usually
considered a good idea to call $lh->SUPER::init in it (presumably at the
beginning), so that all classes get a chance to initialize a new object
however they see fit.

=item *

YourProjClass->fallback_languages()

C<get_handle> appends the return value of this to the end of
whatever list of languages you pass C<get_handle>.  Unless
you override this method, your project class
will inherit Locale::Maketext's C<fallback_languages>, which
currently returns C<('i-default', 'en', 'en-US')>.
("i-default" is defined in RFC 2277).

This method (by having it return the name
of a language-tag that has an existing language class)
can be used for making sure that
C<get_handle> will always manage to construct a language
handle (assuming your language classes are in an appropriate
@INC directory).  Or you can use the next method:

=item *

YourProjClass->fallback_language_classes()

C<get_handle> appends the return value of this to the end
of the list of classes it will try using.  Unless
you override this method, your project class
will inherit Locale::Maketext's C<fallback_language_classes>,
which currently returns an empty list, C<()>.
By setting this to some value (namely, the name of a loadable
language class), you can be sure that
C<get_handle> will always manage to construct a language
handle.

=back

=head2 The "maketext" Method

This is the most important method in Locale::Maketext:

    $text = $lh->maketext(I<key>, ...parameters for this phrase...);

This looks in the %Lexicon of the language handle
$lh and all its superclasses, looking
for an entry whose key is the string I<key>.  Assuming such
an entry is found, various things then happen, depending on the
value found:

If the value is a scalarref, the scalar is dereferenced and returned
(and any parameters are ignored).

If the value is a coderef, we return &$value($lh, ...parameters...).

If the value is a string that I<doesn't> look like it's in Bracket Notation,
we return it (after replacing it with a scalarref, in its %Lexicon).

If the value I<does> look like it's in Bracket Notation, then we compile
it into a sub, replace the string in the %Lexicon with the new coderef,
and then we return &$new_sub($lh, ...parameters...).

Bracket Notation is discussed in a later section.  Note
that trying to compile a string into Bracket Notation can throw
an exception if the string is not syntactically valid (say, by not
balancing brackets right.)

Also, calling &$coderef($lh, ...parameters...) can throw any sort of
exception (if, say, code in that sub tries to divide by zero).  But
a very common exception occurs when you have Bracket
Notation text that says to call a method "foo", but there is no such
method.  (E.g., "You have [quaB<tn>,_1,ball]." will throw an exception
on trying to call $lh->quaB<tn>($_[1],'ball') -- you presumably meant
"quant".)  C<maketext> catches these exceptions, but only to make the
error message more readable, at which point it rethrows the exception.

An exception I<may> be thrown if I<key> is not found in any
of $lh's %Lexicon hashes.  What happens if a key is not found,
is discussed in a later section, "Controlling Lookup Failure".

Note that you might find it useful in some cases to override
the C<maketext> method with an "after method", if you want to
translate encodings, or even scripts:

    package YrProj::zh_cn; # Chinese with PRC-style glyphs
    use base ('YrProj::zh_tw');  # Taiwan-style
    sub maketext {
      my $self = shift(@_);
      my $value = $self->maketext(@_);
      return Chineeze::taiwan2mainland($value);
    }

Or you may want to override it with something that traps
any exceptions, if that's critical to your program:

  sub maketext {
    my($lh, @stuff) = @_;
    my $out;
    eval { $out = $lh->SUPER::maketext(@stuff) };
    return $out unless $@;
    ...otherwise deal with the exception...
  }

Other than those two situations, I don't imagine that
it's useful to override the C<maketext> method.  (If
you run into a situation where it is useful, I'd be
interested in hearing about it.)

=over

=item $lh->fail_with I<or> $lh->fail_with(I<PARAM>)

=item $lh->failure_handler_auto

These two methods are discussed in the section "Controlling
Lookup Failure".

=back

=head2 Utility Methods

These are methods that you may find it handy to use, generally
from %Lexicon routines of yours (whether expressed as
Bracket Notation or not).

=over

=item $language->quant($number, $singular)

=item $language->quant($number, $singular, $plural)

=item $language->quant($number, $singular, $plural, $negative)

This is generally meant to be called from inside Bracket Notation
(which is discussed later), as in 

     "Your search matched [quant,_1,document]!"

It's for I<quantifying> a noun (i.e., saying how much of it there is,
while giving the correct form of it).  The behavior of this method is
handy for English and a few other Western European languages, and you
should override it for languages where it's not suitable.  You can feel
free to read the source, but the current implementation is basically
as this pseudocode describes:

     if $number is 0 and there's a $negative,
        return $negative;
     elsif $number is 1,
        return "1 $singular";
     elsif there's a $plural,
        return "$number $plural";
     else
        return "$number " . $singular . "s";
     #
     # ...except that we actually call numf to
     #  stringify $number before returning it.

So for English (with Bracket Notation)
C<"...[quant,_1,file]..."> is fine (for 0 it returns "0 files",
for 1 it returns "1 file", and for more it returns "2 files", etc.)

But for "directory", you'd want C<"[quant,_1,directory,directories]">
so that our elementary C<quant> method doesn't think that the
plural of "directory" is "directorys".  And you might find that the
output may sound better if you specify a negative form, as in:

     "[quant,_1,file,files,No files] matched your query.\n"

Remember to keep in mind verb agreement (or adjectives too, in
other languages), as in:

     "[quant,_1,document] were matched.\n"

Because if _1 is one, you get "1 document B<were> matched".
An acceptable hack here is to do something like this:

     "[quant,_1,document was, documents were] matched.\n"

=item $language->numf($number)

This returns the given number formatted nicely according to
this language's conventions.  Maketext's default method is
mostly to just take the normal string form of the number
(applying sprintf "%G" for only very large numbers), and then
to add commas as necessary.  (Except that
we apply C<tr/,./.,/> if $language->{'numf_comma'} is true;
that's a bit of a hack that's useful for languages that express
two million as "2.000.000" and not as "2,000,000").

If you want anything fancier, consider overriding this with something
that uses L<Number::Format|Number::Format>, or does something else
entirely.

Note that numf is called by quant for stringifying all quantifying
numbers.

=item $language->sprintf($format, @items)

This is just a wrapper around Perl's normal C<sprintf> function.
It's provided so that you can use "sprintf" in Bracket Notation:

     "Couldn't access datanode [sprintf,%10x=~[%s~],_1,_2]!\n"

returning...

     Couldn't access datanode      Stuff=[thangamabob]!

=item $language->language_tag()

Currently this just takes the last bit of C<ref($language)>, turns
underscores to dashes, and returns it.  So if $language is
an object of class Hee::HOO::Haw::en_us, $language->language_tag()
returns "en-us".  (Yes, the usual representation for that language
tag is "en-US", but case is I<never> considered meaningful in
language-tag comparison.)

You may override this as you like; Maketext doesn't use it for
anything.

=item $language->encoding()

Currently this isn't used for anything, but it's provided
(with default value of
C<(ref($language) && $language-E<gt>{'encoding'})) or "iso-8859-1">
) as a sort of suggestion that it may be useful/necessary to
associate encodings with your language handles (whether on a
per-class or even per-handle basis.)

=back

=head2 Language Handle Attributes and Internals

A language handle is a flyweight object -- i.e., it doesn't (necessarily)
carry any data of interest, other than just being a member of
whatever class it belongs to.

A language handle is implemented as a blessed hash.  Subclasses of yours
can store whatever data you want in the hash.  Currently the only hash
entry used by any crucial Maketext method is "fail", so feel free to
use anything else as you like.

B<Remember: Don't be afraid to read the Maketext source if there's
any point on which this documentation is unclear.>  This documentation
is vastly longer than the module source itself.

=over

=back

=head1 LANGUAGE CLASS HIERARCHIES

These are Locale::Maketext's assumptions about the class
hierarchy formed by all your language classes:

=over

=item *

You must have a project base class, which you load, and
which you then use as the first argument in
the call to YourProjClass->get_handle(...).  It should derive
(whether directly or indirectly) from Locale::Maketext.
It B<doesn't matter> how you name this class, although assuming this
is the localization component of your Super Mega Program,
good names for your project class might be
SuperMegaProgram::Localization, SuperMegaProgram::L10N,
SuperMegaProgram::I18N, SuperMegaProgram::International,
or even SuperMegaProgram::Languages or SuperMegaProgram::Messages.

=item *

Language classes are what YourProjClass->get_handle will try to load.
It will look for them by taking each language-tag (B<skipping> it
if it doesn't look like a language-tag or locale-tag!), turning it to
all lowercase, turning dashes to underscores, and appending it
to YourProjClass . "::".  So this:

  $lh = YourProjClass->get_handle(
    'en-US', 'fr', 'kon', 'i-klingon', 'i-klingon-romanized'
  );

will try loading the classes 
YourProjClass::en_us (note lowercase!), YourProjClass::fr, 
YourProjClass::kon,
YourProjClass::i_klingon
and YourProjClass::i_klingon_romanized.  (And it'll stop at the
first one that actually loads.)

=item *

I assume that each language class derives (directly or indirectly)
from your project class, and also defines its @ISA, its %Lexicon,
or both.  But I anticipate no dire consequences if these assumptions
do not hold.

=item *

Language classes may derive from other language classes (although they
should have "use I<Thatclassname>" or "use base qw(I<...classes...>)").
They may derive from the project
class.  They may derive from some other class altogether.  Or via
multiple inheritance, it may derive from any mixture of these.

=item *

I foresee no problems with having multiple inheritance in
your hierarchy of language classes.  (As usual, however, Perl will
complain bitterly if you have a cycle in the hierarchy: i.e., if
any class is its own ancestor.)

=back

=head1 ENTRIES IN EACH LEXICON

A typical %Lexicon entry is meant to signify a phrase,
taking some number (0 or more) of parameters.  An entry
is meant to be accessed by via
a string I<key> in $lh->maketext(I<key>, ...parameters...),
which should return a string that is generally meant for
be used for "output" to the user -- regardless of whether
this actually means printing to STDOUT, writing to a file,
or putting into a GUI widget.

While the key must be a string value (since that's a basic
restriction that Perl places on hash keys), the value in
the lexicon can currently be of several types:
a defined scalar, scalarref, or coderef.  The use of these is
explained above, in the section 'The "maketext" Method', and
Bracket Notation for strings is discussed in the next section.

While you can use arbitrary unique IDs for lexicon keys
(like "_min_larger_max_error"), it is often
useful for if an entry's key is itself a valid value, like
this example error message:

  "Minimum ([_1]) is larger than maximum ([_2])!\n",

Compare this code that uses an arbitrary ID...

  die $lh->maketext( "_min_larger_max_error", $min, $max )
   if $min > $max;

...to this code that uses a key-as-value:

  die $lh->maketext(
   "Minimum ([_1]) is larger than maximum ([_2])!\n",
   $min, $max
  ) if $min > $max;

The second is, in short, more readable.  In particular, it's obvious
that the number of parameters you're feeding to that phrase (two) is
the number of parameters that it I<wants> to be fed.  (Since you see
_1 and a _2 being used in the key there.)

Also, once a project is otherwise
complete and you start to localize it, you can scrape together
all the various keys you use, and pass it to a translator; and then
the translator's work will go faster if what he's presented is this:

 "Minimum ([_1]) is larger than maximum ([_2])!\n",
  => "",   # fill in something here, Jacques!

rather than this more cryptic mess:

 "_min_larger_max_error"
  => "",   # fill in something here, Jacques

I think that keys as lexicon values makes the completed lexicon
entries more readable:

 "Minimum ([_1]) is larger than maximum ([_2])!\n",
  => "Le minimum ([_1]) est plus grand que le maximum ([_2])!\n",

Also, having valid values as keys becomes very useful if you set
up an _AUTO lexicon.  _AUTO lexicons are discussed in a later
section.

I almost always use keys that are themselves
valid lexicon values.  One notable exception is when the value is
quite long.  For example, to get the screenful of data that
a command-line program might return when given an unknown switch,
I often just use a brief, self-explanatory key such as "_USAGE_MESSAGE".  At that point I then go
and immediately to define that lexicon entry in the
ProjectClass::L10N::en lexicon (since English is always my "project
language"):

  '_USAGE_MESSAGE' => <<'EOSTUFF',
  ...long long message...
  EOSTUFF

and then I can use it as:

  getopt('oDI', \%opts) or die $lh->maketext('_USAGE_MESSAGE');

Incidentally,
note that each class's C<%Lexicon> inherits-and-extends
the lexicons in its superclasses.  This is not because these are
special hashes I<per se>, but because you access them via the
C<maketext> method, which looks for entries across all the
C<%Lexicon> hashes in a language class I<and> all its ancestor classes.
(This is because the idea of "class data" isn't directly implemented
in Perl, but is instead left to individual class-systems to implement
as they see fit..)

Note that you may have things stored in a lexicon
besides just phrases for output:  for example, if your program
takes input from the keyboard, asking a "(Y/N)" question,
you probably need to know what the equivalent of "Y[es]/N[o]" is
in whatever language.  You probably also need to know what
the equivalents of the answers "y" and "n" are.  You can
store that information in the lexicon (say, under the keys
"~answer_y" and "~answer_n", and the long forms as
"~answer_yes" and "~answer_no", where "~" is just an ad-hoc
character meant to indicate to programmers/translators that
these are not phrases for output).

Or instead of storing this in the language class's lexicon,
you can (and, in some cases, really should) represent the same bit
of knowledge as code in a method in the language class.  (That
leaves a tidy distinction between the lexicon as the things we
know how to I<say>, and the rest of the things in the lexicon class
as things that we know how to I<do>.)  Consider
this example of a processor for responses to French "oui/non"
questions:

  sub y_or_n {
    return undef unless defined $_[1] and length $_[1];
    my $answer = lc $_[1];  # smash case
    return 1 if $answer eq 'o' or $answer eq 'oui';
    return 0 if $answer eq 'n' or $answer eq 'non';
    return undef;
  }

...which you'd then call in a construct like this:

  my $response;
  until(defined $response) {
    print $lh->maketext("Open the pod bay door (y/n)? ");
    $response = $lh->y_or_n( get_input_from_keyboard_somehow() );
  }
  if($response) { $pod_bay_door->open()         }
  else          { $pod_bay_door->leave_closed() }

Other data worth storing in a lexicon might be things like
filenames for language-targetted resources:

  ...
  "_main_splash_png"
    => "/styles/en_us/main_splash.png",
  "_main_splash_imagemap"
    => "/styles/en_us/main_splash.incl",
  "_general_graphics_path"
    => "/styles/en_us/",
  "_alert_sound"
    => "/styles/en_us/hey_there.wav",
  "_forward_icon"
   => "left_arrow.png",
  "_backward_icon"
   => "right_arrow.png",
  # In some other languages, left equals
  #  BACKwards, and right is FOREwards.
  ...

You might want to do the same thing for expressing key bindings
or the like (since hardwiring "q" as the binding for the function
that quits a screen/menu/program is useful only if your language
happens to associate "q" with "quit"!)

=head1 BRACKET NOTATION

Bracket Notation is a crucial feature of Locale::Maketext.  I mean
Bracket Notation to provide a replacement for the use of sprintf formatting.
Everything you do with Bracket Notation could be done with a sub block,
but bracket notation is meant to be much more concise.

Bracket Notation is a like a miniature "template" system (in the sense
of L<Text::Template|Text::Template>, not in the sense of C++ templates),
where normal text is passed thru basically as is, but text in special
regions is specially interpreted.  In Bracket Notation, you use square brackets ("[...]"),
not curly braces ("{...}") to note sections that are specially interpreted.

For example, here all the areas that are taken literally are underlined with
a "^", and all the in-bracket special regions are underlined with an X:

  "Minimum ([_1]) is larger than maximum ([_2])!\n",
   ^^^^^^^^^ XX ^^^^^^^^^^^^^^^^^^^^^^^^^^ XX ^^^^

When that string is compiled from bracket notation into a real Perl sub,
it's basically turned into:

  sub {
    my $lh = $_[0];
    my @params = @_;
    return join '',
      "Minimum (",
      ...some code here...
      ") is larger than maximum (",
      ...some code here...
      ")!\n",
  }
  # to be called by $lh->maketext(KEY, params...)
   
In other words, text outside bracket groups is turned into string
literals.  Text in brackets is rather more complex, and currently follows
these rules:

=over

=item *

Bracket groups that are empty, or which consist only of whitespace,
are ignored.  (Examples: "[]", "[    ]", or a [ and a ] with returns
and/or tabs and/or spaces between them.

Otherwise, each group is taken to be a comma-separated group of items,
and each item is interpreted as follows:

=item *

An item that is "_I<digits>" or "_-I<digits>" is interpreted as
$_[I<value>].  I.e., "_1" becomes with $_[1], and "_-3" is interpreted
as $_[-3] (in which case @_ should have at least three elements in it).
Note that $_[0] is the language handle, and is typically not named
directly.

=item *

An item "_*" is interpreted to mean "all of @_ except $_[0]".
I.e., C<@_[1..$#_]>.  Note that this is an empty list in the case
of calls like $lh->maketext(I<key>) where there are no
parameters (except $_[0], the language handle).

=item *

Otherwise, each item is interpreted as a string literal.

=back

The group as a whole is interpreted as follows:

=over

=item *

If the first item in a bracket group looks like a method name,
then that group is interpreted like this:

  $lh->that_method_name(
    ...rest of items in this group...
  ),

=item *

If the first item in a bracket group is "*", it's taken as shorthand
for the so commonly called "quant" method.  Similarly, if the first
item in a bracket group is "#", it's taken to be shorthand for
"numf".

=item *

If the first item in a bracket group is the empty-string, or "_*"
or "_I<digits>" or "_-I<digits>", then that group is interpreted
as just the interpolation of all its items:

  join('',
    ...rest of items in this group...
  ),

Examples:  "[_1]" and "[,_1]", which are synonymous; and
"C<[,ID-(,_4,-,_2,)]>", which compiles as
C<join "", "ID-(", $_[4], "-", $_[2], ")">.

=item *

Otherwise this bracket group is invalid.  For example, in the group
"[!@#,whatever]", the first item C<"!@#"> is neither the empty-string,
"_I<number>", "_-I<number>", "_*", nor a valid method name; and so
Locale::Maketext will throw an exception of you try compiling an
expression containing this bracket group.

=back

Note, incidentally, that items in each group are comma-separated,
not C</\s*,\s*/>-separated.  That is, you might expect that this
bracket group:

  "Hoohah [foo, _1 , bar ,baz]!"

would compile to this:

  sub {
    my $lh = $_[0];
    return join '',
      "Hoohah ",
      $lh->foo( $_[1], "bar", "baz"),
      "!",
  }

But it actually compiles as this:

  sub {
    my $lh = $_[0];
    return join '',
      "Hoohah ",
      $lh->foo(" _1 ", " bar ", "baz"),  # note the <space> in " bar "
      "!",
  }

In the notation discussed so far, the characters "[" and "]" are given
special meaning, for opening and closing bracket groups, and "," has
a special meaning inside bracket groups, where it separates items in the
group.  This begs the question of how you'd express a literal "[" or
"]" in a Bracket Notation string, and how you'd express a literal
comma inside a bracket group.  For this purpose I've adopted "~" (tilde)
as an escape character:  "~[" means a literal '[' character anywhere
in Bracket Notation (i.e., regardless of whether you're in a bracket
group or not), and ditto for "~]" meaning a literal ']', and "~," meaning
a literal comma.  (Altho "," means a literal comma outside of
bracket groups -- it's only inside bracket groups that commas are special.)

And on the off chance you need a literal tilde in a bracket expression,
you get it with "~~".

Currently, an unescaped "~" before a character
other than a bracket or a comma is taken to mean just a "~" and that
character.  I.e., "~X" means the same as "~~X" -- i.e., one literal tilde,
and then one literal "X".  However, by using "~X", you are assuming that
no future version of Maketext will use "~X" as a magic escape sequence.
In practice this is not a great problem, since first off you can just
write "~~X" and not worry about it; second off, I doubt I'll add lots
of new magic characters to bracket notation; and third off, you
aren't likely to want literal "~" characters in your messages anyway,
since it's not a character with wide use in natural language text.

Brackets must be balanced -- every openbracket must have
one matching closebracket, and vice versa.  So these are all B<invalid>:

  "I ate [quant,_1,rhubarb pie."
  "I ate [quant,_1,rhubarb pie[."
  "I ate quant,_1,rhubarb pie]."
  "I ate quant,_1,rhubarb pie[."

Currently, bracket groups do not nest.  That is, you B<cannot> say:

  "Foo [bar,baz,[quux,quuux]]\n";

If you need a notation that's that powerful, use normal Perl:

  %Lexicon = (
    ...
    "some_key" => sub {
      my $lh = $_[0];
      join '',
        "Foo ",
        $lh->bar('baz', $lh->quux('quuux')),
        "\n",
    },
    ...
  );

Or write the "bar" method so you don't need to pass it the
output from calling quux.

I do not anticipate that you will need (or particularly want)
to nest bracket groups, but you are welcome to email me with
convincing (real-life) arguments to the contrary.

=head1 AUTO LEXICONS

If maketext goes to look in an individual %Lexicon for an entry
for I<key> (where I<key> does not start with an underscore), and
sees none, B<but does see> an entry of "_AUTO" => I<some_true_value>,
then we actually define $Lexicon{I<key>} = I<key> right then and there,
and then use that value as if it had been there all
along.  This happens before we even look in any superclass %Lexicons!

(This is meant to be somewhat like the AUTOLOAD mechanism in
Perl's function call system -- or, looked at another way,
like the L<AutoLoader|AutoLoader> module.)

I can picture all sorts of circumstances where you just
do not want lookup to be able to fail (since failing
normally means that maketext throws a C<die>, although
see the next section for greater control over that).  But
here's one circumstance where _AUTO lexicons are meant to
be I<especially> useful:

As you're writing an application, you decide as you go what messages
you need to emit.  Normally you'd go to write this:

  if(-e $filename) {
    go_process_file($filename)
  } else {
    print qq{Couldn't find file "$filename"!\n};
  }

but since you anticipate localizing this, you write:

  use ThisProject::I18N;
  my $lh = ThisProject::I18N->get_handle();
   # For the moment, assume that things are set up so
   # that we load class ThisProject::I18N::en
   # and that that's the class that $lh belongs to.
  ...
  if(-e $filename) {
    go_process_file($filename)
  } else {
    print $lh->maketext(
      qq{Couldn't find file "[_1]"!\n}, $filename
    );
  }

Now, right after you've just written the above lines, you'd
normally have to go open the file 
ThisProject/I18N/en.pm, and immediately add an entry:

  "Couldn't find file \"[_1]\"!\n"
  => "Couldn't find file \"[_1]\"!\n",

But I consider that somewhat of a distraction from the work
of getting the main code working -- to say nothing of the fact
that I often have to play with the program a few times before
I can decide exactly what wording I want in the messages (which
in this case would require me to go changing three lines of code:
the call to maketext with that key, and then the two lines in
ThisProject/I18N/en.pm).

However, if you set "_AUTO => 1" in the %Lexicon in,
ThisProject/I18N/en.pm (assuming that English (en) is
the language that all your programmers will be using for this
project's internal message keys), then you don't ever have to
go adding lines like this

  "Couldn't find file \"[_1]\"!\n"
  => "Couldn't find file \"[_1]\"!\n",

to ThisProject/I18N/en.pm, because if _AUTO is true there,
then just looking for an entry with the key "Couldn't find
file \"[_1]\"!\n" in that lexicon will cause it to be added,
with that value!

Note that the reason that keys that start with "_"
are immune to _AUTO isn't anything generally magical about
the underscore character -- I just wanted a way to have most
lexicon keys be autoable, except for possibly a few, and I
arbitrarily decided to use a leading underscore as a signal
to distinguish those few.

=head1 CONTROLLING LOOKUP FAILURE

If you call $lh->maketext(I<key>, ...parameters...),
and there's no entry I<key> in $lh's class's %Lexicon, nor
in the superclass %Lexicon hash, I<and> if we can't auto-make
I<key> (because either it starts with a "_", or because none
of its lexicons have C<_AUTO =E<gt> 1,>), then we have
failed to find a normal way to maketext I<key>.  What then
happens in these failure conditions, depends on the $lh object's
"fail" attribute.

If the language handle has no "fail" attribute, maketext
will simply throw an exception (i.e., it calls C<die>, mentioning
the I<key> whose lookup failed, and naming the line number where
the calling $lh->maketext(I<key>,...) was.

If the language handle has a "fail" attribute whose value is a
coderef, then $lh->maketext(I<key>,...params...) gives up and calls:

  return $that_subref->($lh, $key, @params);

Otherwise, the "fail" attribute's value should be a string denoting
a method name, so that $lh->maketext(I<key>,...params...) can
give up with:

  return $lh->$that_method_name($phrase, @params);

The "fail" attribute can be accessed with the C<fail_with> method:

  # Set to a coderef:
  $lh->fail_with( \&failure_handler );

  # Set to a method name:
  $lh->fail_with( 'failure_method' );
  
  # Set to nothing (i.e., so failure throws a plain exception)
  $lh->fail_with( undef );
  
  # Get the current value
  $handler = $lh->fail_with();

Now, as to what you may want to do with these handlers:  Maybe you'd
want to log what key failed for what class, and then die.  Maybe
you don't like C<die> and instead you want to send the error message
to STDOUT (or wherever) and then merely C<exit()>.

Or maybe you don't want to C<die> at all!  Maybe you could use a
handler like this:

  # Make all lookups fall back onto an English value,
  #  but only after we log it for later fingerpointing.
  my $lh_backup = ThisProject->get_handle('en');
  open(LEX_FAIL_LOG, ">>wherever/lex.log") || die "GNAARGH $!";
  sub lex_fail {
    my($failing_lh, $key, $params) = @_;
    print LEX_FAIL_LOG scalar(localtime), "\t",
       ref($failing_lh), "\t", $key, "\n";
    return $lh_backup->maketext($key,@params);
  }

Some users have expressed that they think this whole mechanism of
having a "fail" attribute at all, seems a rather pointless complication.
But I want Locale::Maketext to be usable for software projects of I<any>
scale and type; and different software projects have different ideas
of what the right thing is to do in failure conditions.  I could simply
say that failure always throws an exception, and that if you want to be
careful, you'll just have to wrap every call to $lh->maketext in an
S<eval { }>.  However, I want programmers to reserve the right (via
the "fail" attribute) to treat lookup failure as something other than
an exception of the same level of severity as a config file being
unreadable, or some essential resource being inaccessible.

One possibly useful value for the "fail" attribute is the method name
"failure_handler_auto".  This is a method defined in the class
Locale::Maketext itself.  You set it with:

  $lh->fail_with('failure_handler_auto');

Then when you call $lh->maketext(I<key>, ...parameters...) and
there's no I<key> in any of those lexicons, maketext gives up with

  return $lh->failure_handler_auto($key, @params);

But failure_handler_auto, instead of dying or anything, compiles
$key, caching it in

    $lh->{'failure_lex'}{$key} = $complied

and then calls the compiled value, and returns that.  (I.e., if
$key looks like bracket notation, $compiled is a sub, and we return
&{$compiled}(@params); but if $key is just a plain string, we just
return that.)

The effect of using "failure_auto_handler"
is like an AUTO lexicon, except that it 1) compiles $key even if
it starts with "_", and 2) you have a record in the new hashref
$lh->{'failure_lex'} of all the keys that have failed for
this object.  This should avoid your program dying -- as long
as your keys aren't actually invalid as bracket code, and as
long as they don't try calling methods that don't exist.

"failure_auto_handler" may not be exactly what you want, but I
hope it at least shows you that maketext failure can be mitigated
in any number of very flexible ways.  If you can formalize exactly
what you want, you should be able to express that as a failure
handler.  You can even make it default for every object of a given
class, by setting it in that class's init:

  sub init {
    my $lh = $_[0];  # a newborn handle
    $lh->SUPER::init();
    $lh->fail_with('my_clever_failure_handler');
    return;
  }
  sub my_clever_failure_handler {
    ...you clever things here...
  }

=head1 HOW TO USE MAKETEXT

Here is a brief checklist on how to use Maketext to localize
applications:

=over

=item *

Decide what system you'll use for lexicon keys.  If you insist,
you can use opaque IDs (if you're nostalgic for C<catgets>),
but I have better suggestions in the
section "Entries in Each Lexicon", above.  Assuming you opt for
meaningful keys that double as values (like "Minimum ([_1]) is
larger than maximum ([_2])!\n"), you'll have to settle on what
language those should be in.  For the sake of argument, I'll
call this English, specifically American English, "en-US".

=item *

Create a class for your localization project.  This is
the name of the class that you'll use in the idiom:

  use Projname::L10N;
  my $lh = Projname::L10N->get_handle(...) || die "Language?";

Assuming you call your class Projname::L10N, create a class
consisting minimally of:

  package Projname::L10N;
  use base qw(Locale::Maketext);
  ...any methods you might want all your languages to share...
  
  # And, assuming you want the base class to be an _AUTO lexicon,
  # as is discussed a few sections up:
  
  1;

=item *

Create a class for the language your internal keys are in.  Name
the class after the language-tag for that language, in lowercase,
with dashes changed to underscores.  Assuming your project's first
language is US English, you should call this Projname::L10N::en_us.
It should consist minimally of:

  package Projname::L10N::en_us;
  use base qw(Projname::L10N);
  %Lexicon = (
    '_AUTO' => 1,
  );
  1;

(For the rest of this section, I'll assume that this "first
language class" of Projname::L10N::en_us has
_AUTO lexicon.)

=item *

Go and write your program.  Everywhere in your program where 
you would say:

  print "Foobar $thing stuff\n";

instead do it thru maketext, using no variable interpolation in
the key:

  print $lh->maketext("Foobar [_1] stuff\n", $thing);

If you get tired of constantly saying C<print $lh-E<gt>maketext>,
consider making a functional wrapper for it, like so:

  use Projname::L10N;
  use vars qw($lh);
  $lh = Projname::L10N->get_handle(...) || die "Language?";
  sub pmt (@) { print( $lh->maketext(@_)) }
   # "pmt" is short for "Print MakeText"
  $Carp::Verbose = 1;
   # so if maketext fails, we see made the call to pmt

Besides whole phrases meant for output, anything language-dependent
should be put into the class Projname::L10N::en_us,
whether as methods, or as lexicon entries -- this is discussed
in the section "Entries in Each Lexicon", above.

=item *

Once the program is otherwise done, and once its localization for
the first language works right (via the data and methods in
Projname::L10N::en_us), you can get together the data for translation.
If your first language lexicon isn't an _AUTO lexicon, then you already
have all the messages explicitly in the lexicon (or else you'd be
getting exceptions thrown when you call $lh->maketext to get
messages that aren't in there).  But if you were (advisedly) lazy and are
using an _AUTO lexicon, then you've got to make a list of all the phrases
that you've so far been letting _AUTO generate for you.  There are very
many ways to assemble such a list.  The most straightforward is to simply
grep the source for every occurrence of "maketext" (or calls
to wrappers around it, like the above C<pmt> function), and to log the
following phrase.

=item *

You may at this point want to consider whether your base class 
(Projname::L10N), from which all lexicons inherit from (Projname::L10N::en,
Projname::L10N::es, etc.), should be an _AUTO lexicon.  It may be true
that in theory, all needed messages will be in each language class;
but in the presumably unlikely or "impossible" case of lookup failure,
you should consider whether your program should throw an exception,
emit text in English (or whatever your project's first language is),
or some more complex solution as described in the section
"Controlling Lookup Failure", above.

=item *

Submit all messages/phrases/etc. to translators.

(You may, in fact, want to start with localizing to I<one> other language
at first, if you're not sure that you've properly abstracted the
language-dependent parts of your code.)

Translators may request clarification of the situation in which a
particular phrase is found.  For example, in English we are entirely happy
saying "I<n> files found", regardless of whether we mean "I looked for files,
and found I<n> of them" or the rather distinct situation of "I looked for
something else (like lines in files), and along the way I saw I<n>
files."  This may involve rethinking things that you thought quite clear:
should "Edit" on a toolbar be a noun ("editing") or a verb ("to edit")?  Is
there already a conventionalized way to express that menu option, separate
from the target language's normal word for "to edit"?

In all cases where the very common phenomenon of quantification
(saying "I<N> files", for B<any> value of N)
is involved, each translator should make clear what dependencies the
number causes in the sentence.  In many cases, dependency is
limited to words adjacent to the number, in places where you might
expect them ("I found the-?PLURAL I<N>
empty-?PLURAL directory-?PLURAL"), but in some cases there are
unexpected dependencies ("I found-?PLURAL ..."!) as well as long-distance
dependencies "The I<N> directory-?PLURAL could not be deleted-?PLURAL"!).

Remind the translators to consider the case where N is 0:
"0 files found" isn't exactly natural-sounding in any language, but it
may be unacceptable in many -- or it may condition special
kinds of agreement (similar to English "I didN'T find ANY files").

Remember to ask your translators about numeral formatting in their
language, so that you can override the C<numf> method as
appropriate.  Typical variables in number formatting are:  what to
use as a decimal point (comma? period?); what to use as a thousands
separator (space? nonbreaking space? comma? period? small
middot? prime? apostrophe?); and even whether the so-called "thousands
separator" is actually for every third digit -- I've heard reports of
two hundred thousand being expressible as "2,00,000" for some Indian
(Subcontinental) languages, besides the less surprising "S<200 000>",
"200.000", "200,000", and "200'000".  Also, using a set of numeral
glyphs other than the usual ASCII "0"-"9" might be appreciated, as via
C<tr/0-9/\x{0966}-\x{096F}/> for getting digits in Devanagari script
(for Hindi, Konkani, others).

The basic C<quant> method that Locale::Maketext provides should be
good for many languages.  For some languages, it might be useful
to modify it (or its constituent C<numerate> method)
to take a plural form in the two-argument call to C<quant>
(as in "[quant,_1,files]") if
it's all-around easier to infer the singular form from the plural, than
to infer the plural form from the singular.

But for other languages (as is discussed at length
in L<Locale::Maketext::TPJ13|Locale::Maketext::TPJ13>), simple
C<quant>/C<numerify> is not enough.  For the particularly problematic
Slavic languages, what you may need is a method which you provide
with the number, the citation form of the noun to quantify, and
the case and gender that the sentence's syntax projects onto that
noun slot.  The method would then be responsible for determining
what grammatical number that numeral projects onto its noun phrase,
and what case and gender it may override the normal case and gender
with; and then it would look up the noun in a lexicon providing
all needed inflected forms.

=item *

You may also wish to discuss with the translators the question of
how to relate different subforms of the same language tag,
considering how this reacts with C<get_handle>'s treatment of
these.  For example, if a user accepts interfaces in "en, fr", and
you have interfaces available in "en-US" and "fr", what should
they get?  You may wish to resolve this by establishing that "en"
and "en-US" are effectively synonymous, by having one class
zero-derive from the other.

For some languages this issue may never come up (Danish is rarely
expressed as "da-DK", but instead is just "da").  And for other
languages, the whole concept of a "generic" form may verge on
being uselessly vague, particularly for interfaces involving voice
media in forms of Arabic or Chinese.

=item *

Once you've localized your program/site/etc. for all desired
languages, be sure to show the result (whether live, or via
screenshots) to the translators.  Once they approve, make every
effort to have it then checked by at least one other speaker of
that language.  This holds true even when (or especially when) the
translation is done by one of your own programmers.  Some
kinds of systems may be harder to find testers for than others,
depending on the amount of domain-specific jargon and concepts
involved -- it's easier to find people who can tell you whether
they approve of your translation for "delete this message" in an
email-via-Web interface, than to find people who can give you
an informed opinion on your translation for "attribute value"
in an XML query tool's interface.

=back

=head1 SEE ALSO

I recommend reading all of these:

L<Locale::Maketext::TPJ13|Locale::Maketext::TPJ13> -- my I<The Perl
Journal> article about Maketext.  It explains many important concepts
underlying Locale::Maketext's design, and some insight into why
Maketext is better than the plain old approach of having 
message catalogs that are just databases of sprintf formats.

L<File::Findgrep|File::Findgrep> is a sample application/module
that uses Locale::Maketext to localize its messages.  For a larger
internationalized system, see also L<Apache::MP3>.

L<I18N::LangTags|I18N::LangTags>.

L<Win32::Locale|Win32::Locale>.

RFC 3066, I<Tags for the Identification of Languages>,
as at http://sunsite.dk/RFC/rfc/rfc3066.html

RFC 2277, I<IETF Policy on Character Sets and Languages>
is at http://sunsite.dk/RFC/rfc/rfc2277.html -- much of it is
just things of interest to protocol designers, but it explains
some basic concepts, like the distinction between locales and
language-tags.

The manual for GNU C<gettext>.  The gettext dist is available in
C<ftp://prep.ai.mit.edu/pub/gnu/> -- get
a recent gettext tarball and look in its "doc/" directory, there's
an easily browsable HTML version in there.  The
gettext documentation asks lots of questions worth thinking
about, even if some of their answers are sometimes wonky,
particularly where they start talking about pluralization.

The Locale/Maketext.pm source.  Obverse that the module is much
shorter than its documentation!

=head1 COPYRIGHT AND DISCLAIMER

Copyright (c) 1999-2004 Sean M. Burke.  All rights reserved.

This library is free software; you can redistribute it and/or modify
it under the same terms as Perl itself.

This program is distributed in the hope that it will be useful, but
without any warranty; without even the implied warranty of
merchantability or fitness for a particular purpose.

=head1 AUTHOR

Sean M. Burke C<sburke@cpan.org>

=cut