xref: /src/contrib/tzcode/tzfile.5 (revision ff2c98b30b57b9763e2a6575f729bab676e6c025)
1.\" This file is in the public domain, so clarified as of
2.\" 1996-06-05 by Arthur David Olson.
3.Dd March 8, 2026
4.Dt TZFILE 5
5.Os
6.Sh NAME
7.Nm tzfile
8.Nd timezone information
9.Sh DESCRIPTION
10The timezone information files used by
11.Xr tzset 3
12are found under
13.Pa /usr/share/zoneinfo .
14These files use the format described in Internet RFC 9636.
15Each file is a sequence of 8-bit bytes.
16In a file, a binary integer is represented by a sequence of one or
17more bytes in network order (bigendian, or high-order byte first),
18with all bits significant,
19a signed binary integer is represented using two's complement,
20and a boolean is represented by a one-byte binary integer that is
21either 0 (false) or 1 (true).
22The format begins with a 44-byte header containing the following fields:
23.Pp
24.Bl -bullet
25.It
26The magic four-byte ASCII sequence
27.Dq "TZif"
28identifies the file as a timezone information file.
29.It
30A byte identifying the version of the file's format
31(as of 2021, either an ASCII NUL,
32.Dq "2" ,
33.Dq "3" ,
34or
35.Dq "4" ) .
36.It
37Fifteen bytes containing zeros reserved for future use.
38.It
39Six four-byte integer values, in the following order:
40.Pp
41.Bl -tag -compat -width tzh_ttisstdcnt
42.It Va tzh_ttisutcnt
43The number of UT/local indicators stored in the file.
44(UT is Universal Time.)
45.It Va tzh_ttisstdcnt
46The number of standard/wall indicators stored in the file.
47.It Va tzh_leapcnt
48The number of leap seconds for which data entries are stored in the file.
49.It Va tzh_timecnt
50The number of transition times for which data entries are stored
51in the file.
52.It Va tzh_typecnt
53The number of local time types for which data entries are stored
54in the file (must not be zero).
55.It Va tzh_charcnt
56The number of bytes of time zone abbreviation strings
57stored in the file.
58.El
59.El
60.Pp
61The above header is followed by the following fields, whose lengths
62depend on the contents of the header:
63.Bl -tag -compat -width tzh_timecnt
64.It Va tzh_timecnt
65four-byte signed integer values sorted in ascending order.
66These values are written in network byte order.
67Each is used as a transition time (as returned by
68.Xt time 2 )
69at which the rules for computing local time change.
70.It Va tzh_timecnt
71one-byte unsigned integer values;
72each one but the last tells which of the different types of local time types
73described in the file is associated with the time period
74starting with the same-indexed transition time
75and continuing up to but not including the next transition time.
76(The last time type is present only for consistency checking with the
77proleptic TZ string described below.)
78These values serve as indices into the next field.
79.It Va tzh_typecnt
80.Vt ttinfo
81entries, each defined as follows:
82.Pp
83.Bd -literal -offset indent
84struct ttinfo {
85	int32_t	tt_utoff;
86	unsigned char	tt_isdst;
87	unsigned char	tt_desigidx;
88};
89.Ed
90.Pp
91Each structure is written as a four-byte signed integer value for
92.Va tt_utoff ,
93in network byte order, followed by a one-byte boolean for
94.Va tt_isdst
95and a one-byte value for
96.Va tt_desigidx .
97In each structure,
98.Va tt_utoff
99gives the number of seconds to be added to UT,
100.Va tt_isdst
101tells whether
102.Va tm_isdst
103should be set by
104.Xr localtime 3
105and
106.Va tt_desigidx
107serves as an index into the array of time zone abbreviation bytes
108that follow the
109.Vt ttinfo
110entries in the file; if the designated string is "\-00", the
111.Vt ttinfo
112entry is a placeholder indicating that local time is unspecified.
113The
114.Va tt_utoff
115value is never equal to \-2**31, to let 32-bit clients negate it without
116overflow.
117Also, in realistic applications
118.Va tt_utoff
119is in the range [\-89999, 93599] (i.e., more than \-25 hours and less
120than 26 hours); this allows easy support by implementations that
121already support the POSIX-required range [\-24:59:59, 25:59:59].
122.It Va tzh_charcnt
123bytes that represent time zone designations,
124which are null-terminated byte strings, each indexed by the
125.Va tt_desigidx
126values mentioned above, and each corresponding to a time zone abbreviation.
127The byte strings can overlap if one is a suffix of the other.
128The encoding of these strings is not specified.
129.It Va tzh_leapcnt
130pairs of four-byte values, written in network byte order;
131the first value of each pair gives the non-negative time
132(as returned by
133.Xr time 3 )
134at which a leap second occurs or at which the leap second table expires;
135the second is a signed integer specifying the correction, which is the
136.Em total
137number of leap seconds to be applied during the time period
138starting at the given time.
139The pairs of values are sorted in strictly ascending order by time.
140Each pair denotes one leap second, either positive or negative,
141except that if the last pair has the same correction as the previous one,
142the last pair denotes the leap second table's expiration time.
143Each leap second is at the end of a UTC calendar month.
144The first leap second has a non-negative occurrence time,
145and is a positive leap second if and only if its correction is positive;
146the correction for each leap second after the first differs
147from the previous leap second by either 1 for a positive leap second,
148or \-1 for a negative leap second.
149If the leap second table is empty, the leap-second correction is zero
150for all timestamps;
151otherwise, for timestamps before the first occurrence time,
152the leap-second correction is zero if the first pair's correction is 1 or \-1,
153and is unspecified otherwise (which can happen only in files
154truncated at the start).
155.It Va tzh_ttisstdcnt
156standard/wall indicators, each stored as a one-byte boolean;
157they tell whether the transition times associated with local time types
158were specified as standard time or local (wall clock) time.
159.It Va tzh_ttisutcnt
160UT/local indicators, each stored as a one-byte boolean;
161they tell whether the transition times associated with local time types
162were specified as UT or local time.
163If a UT/local indicator is set, the corresponding standard/wall indicator
164must also be set.
165.El
166.Pp
167The standard/wall and UT/local indicators were designed for
168transforming a TZif file's transition times into transitions appropriate
169for another time zone specified via
170a proleptic TZ string that lacks rules.
171For example, when TZ="EET\-2EEST" and there is no TZif file "EET\-2EEST",
172the idea was to adapt the transition times from a TZif file with the
173well-known name "posixrules" that was present only for this purpose and
174was a copy of the file "Europe/Brussels", a file with a different UT offset.
175POSIX does not specify the details of this obsolete transformational behavior,
176the default rules are installation-dependent, and no implementation
177is known to support this feature for timestamps past 2037,
178so users desiring (say) Greek time should instead specify
179TZ="Europe/Athens" for better historical coverage, falling back on
180TZ="EET\-2EEST,M3.5.0/3,M10.5.0/4"
181if conformance to POSIX.1-2017 or earlier is required
182and older timestamps need not be handled accurately.
183.Pp
184The
185.Xr localtime 3
186function
187normally uses the first
188.Vt ttinfo
189structure in the file
190if either
191.Va tzh_timecnt
192is zero or the time argument is less than the first transition time recorded
193in the file.
194.Ss Version 2 format
195For version-2-format timezone files,
196the above header and data are followed by a second header and data,
197identical in format except that
198eight bytes are used for each transition time or leap second time.
199(Leap second counts remain four bytes.)
200After the second header and data comes a newline-enclosed string
201in the style of the contents of a proleptic TZ,
202for use in handling instants
203after the last transition time stored in the file
204or for all instants if the file has no transitions.
205The TZ string is empty (i.e., nothing between the newlines)
206if there is no proleptic representation for such instants.
207.PP
208If non-empty, the TZ string must agree with the local time
209type after the last transition time if present in the eight-byte data;
210for example, given the string
211.Dq "WET0WEST,M3.5.0/1,M10.5.0"
212then if a last transition time is in July, the transition's local time
213type must specify a daylight-saving time abbreviated
214.Dq "WEST"
215that is one hour east of UT.
216.PP
217The TZ string can contain time zone abbreviations and UT offsets that
218do not appear elsewhere in the TZif file.
219.PP
220Also, if there is at least one transition, time type 0 is associated
221with the time period from the indefinite past up to but not including
222the earliest transition time.
223.Ss Version 3 format
224For version-3-format timezone files, a TZ string (see
225.Xr newtzset 3 )
226may use the following POSIX.1-2024 extensions to POSIX.1-2017:
227First, as in TZ="<\-02>2<\-01>,M3.5.0/\-1,M10.5.0/0",
228the hours part of its transition times may be signed and range from
229\-167 through 167 instead of being limited to unsigned values
230from 0 through 24.
231Second, as in TZ="XXX3EDT4,0/0,J365/23", DST is in effect all year if it starts
232January 1 at 00:00 and ends December 31 at 24:00 plus the difference
233between daylight saving and standard time.
234.Ss Version 4 format
235For version-4-format TZif files,
236the first leap second record can have a correction that is neither
237+1 nor \-1, to represent truncation of the TZif file at the start.
238Also, if two or more leap second transitions are present and the last
239entry's correction equals the previous one, the last entry
240denotes the expiration of the leap second table instead of a leap second;
241timestamps after this expiration are unreliable in that future
242releases will likely add leap second entries after the expiration, and
243the added leap seconds will change how post-expiration timestamps are treated.
244.Ss Interoperability considerations
245Future changes to the format may append more data.
246.Pp
247Version 1 files are considered a legacy format and
248should not be generated, as they do not support transition
249times after the year 2038.
250Readers that understand only Version 1 must ignore
251any data that extends beyond the calculated end of the version
2521 data block.
253.Pp
254Other than version 1, writers should generate
255the lowest version number needed by a file's data.
256For example, a writer should generate a version 4 file
257only if its leap second table either expires or is truncated at the start.
258Likewise, a writer not generating a version 4 file
259should generate a version 3 file only if
260TZ string extensions are necessary to accurately
261model transition times.
262.Pp
263The sequence of time changes defined by the version 1
264header and data block should be a contiguous sub-sequence
265of the time changes defined by the version 2+ header and data
266block, and by the footer.
267This guideline helps obsolescent version 1 readers
268agree with current readers about timestamps within the
269contiguous sub-sequence.
270It also lets writers not
271supporting obsolescent readers use a
272.Va tzh_timecnt
273of zero
274in the version 1 data block to save space.
275.Pp
276When a TZif file contains a leap second table expiration
277time, TZif readers should either refuse to process
278post-expiration timestamps, or process them as if the expiration
279time did not exist (possibly with an error indication).
280.Pp
281Time zone abbreviations should consist of at least three (3)
282and no more than six (6) ASCII characters from the set of
283alphanumerics,
284.Dq "\-" ,
285and
286.Dq "+" .
287This is for compatibility with POSIX requirements for
288time zone abbreviations.
289.Pp
290A numeric time zone abbreviation should match the UT offset.
291For example, "+0530" should be used only if the UT offset is 5.5 hours
292ahead of UT, and "\-00" should be used only if the UT offset is zero.
293.Pp
294When reading a version 2 or higher file, readers
295should ignore the version 1 header and data block except for
296the purpose of skipping over them.
297.Pp
298Readers should calculate the total lengths of the
299headers and data blocks and check that they all fit within
300the actual file size, as part of a validity check for the file.
301.Pp
302When a positive leap second occurs, readers should append an extra
303second to the local minute containing the second just before the leap
304second.
305If this occurs when the UTC offset is not a multiple of 60
306seconds, the leap second occurs earlier than the last second of the
307local minute and the minute's remaining local seconds are numbered
308through 60 instead of the usual 59; the UTC offset is unaffected.
309.Ss Common interoperability issues
310This section documents common problems in reading or writing TZif files.
311Most of these are problems in generating TZif files for use by
312older readers.
313The goals of this section are to help:
314.Bl -bullet
315.It
316TZif writers output files that avoid common
317pitfalls in older or buggy TZif readers,
318.It
319TZif readers avoid common pitfalls when reading
320files generated by future TZif writers, and
321.It
322any future specification authors see what sort of
323problems arise when the TZif format is changed.
324.El
325.Pp
326When new versions of the TZif format have been defined, a
327design goal has been that a reader can successfully use a TZif
328file even if the file is of a later TZif version than what the
329reader was designed for.
330When complete compatibility was not achieved, an attempt was
331made to limit glitches to rarely used timestamps and allow
332simple partial workarounds in writers designed to generate
333newer-version data useful even for older-version readers.
334This section attempts to document these compatibility issues and
335workarounds as well as documenting other common bugs in
336readers.
337.Pp
338Interoperability problems with TZif include the following:
339.Bl -bullet
340.It
341Some readers examine only version 1 data.
342As a partial workaround, a writer can output as much version 1
343data as possible.
344However, a reader should ignore version 1 data, and should use
345version 2+ data even if the reader's native timestamps have only
34632 bits.
347.It
348Some readers designed for version 2 might mishandle
349timestamps after a version 3 or higher file's last transition, because
350they cannot parse the POSIX.1-2024 extensions to POSIX.1-2017
351in the proleptic TZ string.
352As a partial workaround, a writer can output more transitions
353than necessary, so that only far-future timestamps are
354mishandled by version 2 readers.
355.It
356Some readers might mishandle timestamps after a file's last transition,
357because they require that all abbreviations or UT offsets
358in the proleptic TZ string must also occur somewhere in the file's tables
359of time zone designations and local time type records.
360As a workaround, a writer can output more transitions than necessary,
361so that the other tables contain duplicates of the proleptic TZ string's
362abbreviations and offsets.
363.It
364Some readers designed for version 2 do not support
365permanent daylight saving time with transitions after 24:00
366\(en e.g., a TZ string
367.Dq "EST5EDT,0/0,J365/25"
368denoting permanent Eastern Daylight Time
369(\-04).
370As a workaround, a writer can substitute standard time
371for two time zones east, e.g.,
372.Dq "XXX3EDT4,0/0,J365/23"
373for a time zone with a never-used standard time (XXX, \-03)
374and negative daylight saving time (EDT, \-04) all year.
375Alternatively,
376as a partial workaround, a writer can substitute standard time
377for the next time zone east \(en e.g.,
378.Dq "AST4"
379for permanent
380Atlantic Standard Time (\-04).
381.It
382Some readers designed for version 2 or 3 and that require strict
383conformance to RFC 9636 reject version 4 files whose leap second
384tables are truncated at the start or end in expiration times.
385.It
386Some readers ignore the footer, and instead predict future
387timestamps from the time type of the last transition.
388As a partial workaround, a writer can output more transitions
389than necessary.
390.It
391Some stripped-down readers ignore everything but the footer,
392and use its proleptic TZ string to calculate all timestamps.
393Although this approach often works for current and future timestamps,
394it obviously has problems with past timestamps,
395and even for current timestamps it can fail for settings like
396TZ="Africa/Casablanca".  This corresponds to a TZif file
397containing explicit transitions through the year 2087,
398followed by a footer containing the TZ string
399.Dq <+01>\-1 ,
400which should be used only for timestamps after the last
401explicit transition.
402.It
403Some readers do not use time type 0 for timestamps before
404the first transition, in that they infer a time type using a
405heuristic that does not always select time type 0.
406As a partial workaround, a writer can output a dummy (no-op)
407first transition at an early time.
408.It
409Some readers mishandle timestamps before the first
410transition that has a timestamp that is not less than \-2**31.
411Readers that support only 32-bit timestamps are likely to be
412more prone to this problem, for example, when they process
41364-bit transitions only some of which are representable in 32
414bits.
415As a partial workaround, a writer can output a dummy
416transition at timestamp \-2**31.
417.It
418Some readers mishandle a transition if its timestamp has
419the minimum possible signed 64-bit value.
420Timestamps less than \-2**59 are not recommended.
421.It
422Some readers mishandle proleptic TZ strings that
423contain
424.Dq "<"
425or
426.Dq ">".
427As a partial workaround, a writer can avoid using
428.Dq "<"
429or
430.Dq ">"
431for time zone abbreviations containing only alphabetic
432characters.
433.It
434Many readers mishandle time zone abbreviations that contain
435non-ASCII characters.
436These characters are not recommended.
437.It
438Some readers may mishandle time zone abbreviations that
439contain fewer than 3 or more than 6 characters or that
440contain ASCII characters other than alphanumerics,
441.Dq "\-",
442and
443.Dq "+".
444These abbreviations are not recommended.
445.It
446Some readers mishandle TZif files that specify
447daylight-saving time UT offsets that are less than the UT
448offsets for the corresponding standard time.
449These readers do not support locations like Ireland, which
450uses the equivalent of the TZ string
451.Dq "IST\-1GMT0,M10.5.0,M3.5.0/1",
452observing standard time
453(IST, +01) in summer and daylight saving time (GMT, +00) in winter.
454As a partial workaround, a writer can output data for the
455equivalent of the TZ string
456.Dq "GMT0IST,M3.5.0/1,M10.5.0",
457thus swapping standard and daylight saving time.
458Although this workaround misidentifies which part of the year
459uses daylight saving time, it records UT offsets and time zone
460abbreviations correctly.
461.It
462Some readers generate ambiguous timestamps for positive leap seconds
463that occur when the UTC offset is not a multiple of 60 seconds.
464For example, with UTC offset +01:23:45 and
465a positive leap second 78796801 (1972-06-30 23:59:60 UTC), some readers will
466map both 78796800 and 78796801 to 01:23:45 local time the next day
467instead of mapping the latter to 01:23:46, and they will map 78796815 to
46801:23:59 instead of to 01:23:60.
469This has not yet been a practical problem, since no civil authority
470has observed such UTC offsets since leap seconds were
471introduced in 1972.
472.El
473.Pp
474Some interoperability problems are reader bugs that
475are listed here mostly as warnings to developers of readers.
476.Bl -bullet
477.It
478Some readers do not support negative timestamps.
479Developers of distributed applications should keep this
480in mind if they need to deal with pre-1970 data.
481.It
482Some readers mishandle timestamps before the first
483transition that has a non-negative timestamp.
484Readers that do not support negative timestamps are likely to
485be more prone to this problem.
486.It
487Some readers mishandle time zone abbreviations like
488.Dq "\-08"
489that contain
490.Dq "+",
491.Dq "\-",
492or digits.
493.It
494Some readers mishandle UT offsets that are out of the
495traditional range of \-12 through +12 hours, and so do not
496support locations like Kiritimati that are outside this
497range.
498.It
499Some readers mishandle UT offsets in the range [\-3599, \-1]
500seconds from UT because they integer-divide the offset by
5013600 to get 0 and then display the hour part as
502.Dq "+00" .
503.It
504Some readers mishandle UT offsets that are not a multiple
505of one hour, or of 15 minutes, or of 1 minute.
506.El
507.Sh SEE ALSO
508.Xr time 3 ,
509.Xr localtime 3 ,
510.Xr tzset 3 ,
511.Xr tzsetup 8 ,
512.Xr zic 8 ,
513.Xr zdump 8
514.Rs
515.%A A. Olson
516.%A P. Eggert
517.%A K. Murchison
518.%T "The Time Zone Information Format (TZif)"
519.%R RFC 9636
520.%D October 2024
521.%U https://datatracker.ietf.org/doc/html/rfc9636
522.%U https://doi.org/10.17487/RFC9636
523.Re
524