- tz - lists.iana.org

Re: Addition to Arthur Olsen/4.3bsd table-driven ctime
by Steve Summit April 6, 1987

April 6, 1987

>> I cannot be sure that my customers will install the necessary >> timezone support files. > Neither can you be sure that they will install /etc/passwd. You don't > handle that, why handle this? /etc/passwd is standard on the systems I'm selling to; /etc/zoneinfo isn't (yet). A customer doesn't have to install /etc/passwd to get my software to run; I know he already has it, or else he's got a broken system, and it's his problem, not mine. >> I insist that both timezone and DST corrections be applied, even in >> the absence of /etc/zoneinfo, which means that ctime must carry >> some DST information along in its data segment, so it can perform >> at least as well as it used to. > This is unclear to me. What I'm worried about here is systems that haven't applied any fixes. I detest "two steps forward, one step back" improvements, so I'm trying to arrange that a ctime that I link into my code behave 100% backwards-compatibly when shipped to a site that depends on using the old ctime. By "backwards-compatibly," I mean that any new ctime, when operating in an environment without /etc/zoneinfo, must generate the same results as (i.e. localtime must be functionally identical to) the "old" versions of ctime. If not, there will be periods of time (such as the three-week period when lazy and/or sensible systems simply set their clock ahead by an hour) during which the new version of ctime will generate a different local time value than the rest of the utilities on the system (date, ls, anything linked with the old ctime) when presented with the same kernel timestamp. >> The next question is, should any DST information hard-compiled >> into ctime be fixed to reflect recent changes? The surprising >> answer is "no." > That means that your worst case is off by an hour. (Ignoring double > DST.) The same as 4.3's. What am I missing? No. It's off by 0 hours, at all times of any past or present year, when shipped to a system that applies no fixes. (Actually, there is a worse worst case, that of the site that applies the fixes in Article 13, to arrive at a temporarily fixed but still non table-driven ctime. I can't arrange for things to work correctly in this case without breaking things in the other two cases, and I consider the other two cases -- the no fix case and the Arthur Olsen case -- more likely.) Actually, no one else here at Tektronix is at all concerned about this issue, and the witching hour has come and gone, so it's probably not worth our thrashing this out. I thought my arguments were fairly convincing, and I tried to sweeten the pot by offering working code, but I admit that these are pretty fine points I'm considering, and since nobody else seems to be interested, I think I'll drop the whole thing and go off and find other damsels to rescue :-). Steve Summit

2 1

Ballot 3 -- struct tm
by seismo!nbires!vianet!devine April 4, 1987

April 4, 1987

On this ballot feel free to mix'n'match your choices. 3. How should the structure tm be defined? a. As it is defined in the SVID with no added members. b. Extend tm structure to contain local timezone offset from UT c. Extend tm structure to contain local seasonal time offset from UT d. Extend tm structure to contain timezone name e. Extend tm structure to contain a pointer to the timezone name f. Drop tm_isdst from the structure. Related questions: 1. If the structure is changed, should it be renamed? 2. What does an invalid structure contain? Can there be invalid struct tm's? 3. If the timezone name is included by a char pointer, how does one make a "permanent" copy of the structure? 4. If the timezone name is included, must there be a standard for timezone names so that a different process can make sense out of a passed structure? 5. How is the structure to be used by the mktime() function.

1 0

ballot 2 - result
by seismo!nbires!vianet!devine April 4, 1987

April 4, 1987

Here are the opinions on how the time_t type should be defined. It may not have been clear what I was getting at with the questions about "how does your favorite machine arch allow for char addressing" or "how do you measure distances". Those questions were meant to show the analogous nature of time measurement to distance and to how one selects a particular point. Time_t has been a single unit measure -- seconds. Compare that with how address for Intel chips are made -- segment and offset. In that scheme, there can be multiple "names" for the same location that can be resolved only be a normalization process. Two of the choices (e, f and maybe b) were also composed of two parts that, while they do extend the range, may have drawbacks based on their form. Distances can be given in a form that has several units (miles, feet, inches). Calculations are easier if there is only one unit; no conversions are needed. Bob Devine ------------------------------------ 2. The type time_t should be defined as: a. a signed long integer whose value is the number of seconds since 1/1/70 0:0:0 UT (the SVID definition) b. an implementation defined type (a long, or struct, or float or ?) c. a signed long integer whose value is the number of seconds since 1/1/70 0:0:0 UT but whose bits may have a special, overloaded meaning d. an unsigned long whose value is the number of seconds since 1/1/70 0:0:0 UT and extends to the year 2100 e. a signed long integer whose value is the number of seconds from a system- or implementation-selected date and time f. a companion value to a date_t where time_t is number of seconds since the start of the day represented in date_t --Bob Devine---------- Choose 'a'. It is too late for a change to happen. Using an "overloaded meaning" time_t will bite those who are not aware of the circumstances where a particular bit pattern is not a simple count-of-seconds. The other choices could be supported with a library that parallels the functions provided by time() and ctime(3) funcs. ---------------------- --Mark Horton--------- I vote for c', namely an unsigned long integer whose value is the number of seconds since 1/1/70 0:0:0 UT but whose bits may have a special, overloaded meaning Namely: all 1's means "error" anything whose 25th through 32nd least significant bits (e.g. the 8 high bits in a 32 bit long) are all 1 is interpreted instead as follows: take the low order 24 bits The time represents 0:0:0 (or any unspecified time) on the date that many days since 0/0/0, as a signed 24 bit integer Alternatively, "error" might be taken as 9 high order 1's, a day 2^23 days BC. ---------------------- --Arthur David Olson-- A. If any other definition is decided on, the type name must be something other than time_t. Any other approach will break existing code. ---------------------- --Tom Hofmann--------- I vote for 'f' -- with 'f' we get more independent and are on the "safe side". ---------------------- --Geoff Kuenning------ For me, (a) is the clear choice. Only (a) and (d) provide effective compatibility for most programs. (c) is just a way to add complexity to a programmer's life without adding real power; everybody will have to deal with the special bits, but the people who actually care about them will not be saved any code over using a more general time type. This is contrary to the "spirit of UNIX". More practically, it is contrary to the principle of simplicity that has made UNIX so wonderful for programmers. (b) and (e) cost too much in portability of existing programs (although (e) is not as bad as first appears, since most time calculations are in effect relative). (f) provides too many ways to represent the same time, which means that many programs would have to have code to normalize time representations to the same date_t before they could manipulate data, again violating the principle of simplicity. It seems to me that there are really two problems here: time stamping and arithmetic on stamps, and general operations on points in time. The existing time_t is adequate for the first. Making it signed doesn't really help with the general problem, but does make it reasonable to represent time differences using a time_t. Of the other options, only (b) and (f) address the general problem, and they do so poorly and at the expense of simplicity and portability. I'd rather either design a really easy-to-use general mechanism, or just punt the problem and say "if you want more, use a private data structure". Note that a truly general solution is pretty difficult, since it is easy to argue for both picosecond resolution and megayear range. Since a year is about 10**7 seconds, this requires a dynamic range of 10**25, or about 2**84. Floating-point is a pretty good answer, though of course it introduces all of the myriad problems of floating arithmetic. > A. Is the range of 1970 +/- 67 years sufficient? Yes and no. One can infer from the lifetimes of OS/8 and IBM's DOS that there will be UNIX machines in operation well past the year 2038. However, it is quite likely that none of the machines will be 32-bitters by then, so the question may become moot. The negative range, of course, is pretty useless for serious date work, but it seems pretty adequate for time differences. However, since there will be files like /bin/true that will still have a 1980's modification date in the 21st century, the negative range must always be able to handle the difference between the current time and the epoch. > B. What is a 'long' for machines with different architectures? K&R guarantee that a long has at least 32 bits. Time_t should not be required to be a C "long" type; it should simply be required to be at least 32 bits. The wide-word types of the future can use their 32-bit shorts if they want, or they can use a 128-bit long to increase accuracy. > C. How is an invalid time represented? Why is it necessary to represent one? If you decide on a representation, then everybody has to check for it. This is one thing with IEEE floating point, where the check can be done in parallel hardware, and quite another with a software type. In any case, what is the meaning of the term "invalid time"? All times are valid; they just might not be accurate. ----------------------

1 0

Re: Berkeley variant of time zone stuff
by seismo!sun!guy April 2, 1987

April 2, 1987

I'm sorry, but I'm not following this whole discussion. It seems to me that S5 has the problem here; Ultimately, it does, but unless and until AT&T fixes the problem there needs to be some way of getting around it. S5 has no way (does it?) to do so. You can set "TZ" to the "proper" value. S5R3.1, and I think some of the earlier versions, at least on the 3Bs, have a file "/etc/TIMEZONE" that can be used to find the "proper" value. > That's why I suggested the "defaultzone" routine - you need some > portable way of setting the TZ environment variable; given existing > systems, removing it isn't a portable way of forcing local wall clock > time. To make this portable, you'll have to alter S5, right? To make it portable to *all* S5 versions distributed by AT&T, you would, but to make it portable to the later ones, you would just have to have "defaultzone" grab the value from TZ. We can't fix every version of S5, but we can fix some of them. Note that "alter(ing) S5" consist only of adding a file "/etc/TIMEZONE" whose contents are something like "TZ=PST5PDT". You don't have to alter any code, you just have to add "/etc/TIMEZONE" and the "defaultzone" routine - or throw in your own version of "defaultzone" and #ifdef it. (Perfect portability can't always be achieved; I'm willing to force people to use #ifdefs on occasion if that makes the code "more portable" in the sense that you have to change less to make it run elsewhere.) The notion here would be to add "defaultzone" to the P1003.1 or X3J11 standard (whichever one discusses the TZ environment variable - not that the October 1 X3J11 draft does NOT discuss TZ, so unless they're going to add it P1003.1 had better think about picking it up). Systems with rational ways of handling TZ can just return a null pointer or the string "localtime"; systems that insist that you set TZ to something parsable can snarf the default value from some file and return that. Also note that this discussion is directed at vendors as well as users; a vendor can always add "/etc/TIMEZONE" and "defaultzone" to their system if they have to.

1 0

Re: Berkeley variant of time zone stuff
by seismo!vianet!devine April 2, 1987

April 2, 1987

I just finished reading Keith Bostic's posted patches for 4.[23] time functions. I disagree with the way that a compiled-in table is still used if the TZ env variable is not defined. Such tables should be extirpated. Note that the new 1987 version of when Daylight Saving Time starts **is not a national law**. Each state can decide to use the first or last Sunday in April to start DST. Each state may decide to even have DST. One lesson the should be learned is that DST rules are extremely changeable. The future holds more possible changes: presidential years may have timezone-dependent DST; Colorado may decide to use year-'round DST (the bill was just introduced last week). Will Berkeley issue a new series of code patches every time a change occurs? Bob Devine

1 0

Re: Berkeley variant of time zone stuff
by seismo!berkeley.edu!bostic%okeeffe April 2, 1987

April 2, 1987

> Point taken; clearly the gettimeofday system call must be retained > (which implies that settimeofday has to hang around too) > to help out old binaries. More importantly, it has to be retained for better than second resolution. > Or "/etc/TIMEZONE". S5R3.1 currently has "init" load up the > environment from what's in "/etc/TIMEZONE". Presumably, you could > have "defaultzone" read the setting of TZ from this... > > or generate a string from the abbreviations and other > information stored in the "/etc/zoneinfo/localtime" file. > > or generate that string. > > But there doesn't seem to be any way to get > unsetenv("TZ") > (the Berkeley posting approach) to give the desired effect of > getting your process--and any children that it later forks > off--running on local wall clock time. If you take that approach > on a System V system, and then fork off an "old" binary, the old > binary will end up using GMT. > I'm sorry, but I'm not following this whole discussion. It seems to me that S5 has the problem here; BSD offers two different ways to guarantee that "localtime" is used, setenv("localtime") and unsetenv("TZ"). S5 has no way (does it?) to do so. Adding another way to BSD won't help the situation. Now, if S5 wanted to add something, I could understand wanting to add it to BSD. > That's why I suggested the "defaultzone" routine - you need some > portable way of setting the TZ environment variable; given existing > systems, removing it isn't a portable way of forcing local wall clock > time. To make this portable, you'll have to alter S5, right? --keith

1 0

Re: Addition to Arthur Olsen/4.3bsd table-driven ctime
by Arthur David Olson April 1, 1987

April 1, 1987

> While working over ctime, I came up with a couple of questions: > > In asctime, shouldn't the year really be printed with %4d > (or maybe %-4d) so that the returned string is guaranteed > to have its advertised 26-character length? (I realize > that the code is lifted directly from the X3J11 draft > standard, and that %d will only get it wrong for dates in > the middle ages that a 32-bit time_t can't begin to reach. > On the other hand, asctime gets handed a broken-down tm > struct, so early years are quite possible.) > > Shouldn't the offtime() routine be declared static? > It's not a publicized interface. > > Steve Summit > stevesu(a)copper.tek.com Even %4d won't guarantee the advertised 26-character length if tm_year is big (or small) enough (for example, 10000); bogus tm_wday and tm_mon values can cause problems as well. You may want to send a note to an X3J11 committee member. offtime was declared static in the mod.sources posting unless STD_INSPIRED was defined. Its eventual fate lies more in the hands of standards committees and other folks wiser than yours truly at this point. --ado

1 0

Re: Berkeley variant of time zone stuff
by Arthur David Olson April 1, 1987

April 1, 1987

First a new question, then stuff inspired by old communicaitons. The new: System V systems run on GMT if there's no TZ environment variable. The new time zone stuff arranges to run on whatever's described in the file "/etc/zoneinfo/localtime" if there's no TZ environment variable. Now you *can* force compatibility either by describing GMT in the file "/etc/zoneinfo/localtime" or by not having the file at all (although this plays hell with the localtime(3) claim that you're getting the "best available approximation to local wall clock time). But would it be a better idea--for the sake of compatibility with System V--for the new time zone stuff to use GMT if there's no TZ environment variable? ---------- ado > > Is there a good reason for keeping the gettimeofday (and ado > > settimeofday) system calls? guy > . . .it has to be provided if binary compatibility with guy > old programs is required; they expect the time zone information to be guy > provided by "gettimeofday". It'd be nice to convert those programs, guy > but this isn't always possible or practical. Point taken; clearly the gettimeofday system call must be retained (which implies that settimeofday has to hang around too) to help out old binaries. A slightly ugly possibility is to conditionalize "/usr/include/sys/time.h": struct timezone { #ifdef KERNEL int tz_minuteswest; int tz_dsttime; #else /* !KERNEL */ int :32; int :32; #endif /* !KERNEL */ }; to preclude use of tz_minuteswest and tz_dsttime when a program is recompiled. This is perhaps more encouragement to convert than is appropriate. :-) One advantage, though, is that the change is invisible to gettimeofday callers who are only interested in getting higher precision time. A find /usr/src -name "*.c" -exec grep -l tz_ {} \; is on my list of things to do. ---------- Guy's reminder about supporting old programs also got me thinking about one of Guy's asides: > On systems where TZ *must* be set for proper operation, you would > have to do put it back into the environment. . .One > possibility here might be to have a function > > char *defaultzone(void); /* X3J11 syntax; no arguments */ > > which returns either a pointer to a string indicating the default > time zone, or a NULL pointer if not setting TZ will cause the system > to use the default time zone. (Admittedly, the latter is a frill; on > systems using Arthur's code, for example, it could just as well > return a pointer to a string containing "localtime".) Setting TZ to "localtime" would work fine on a BSD system--"new" programs would be using the new time zone stuff and would honor it; "old" programs would be calling gettimeofday. But if it's a System V system with "old" binaries, "localtime" probably isn't what you want to put in the environment, since doing that would convince such child processes that you wanted to run on GMT with the time zone abbreviation "loc". There are a couple of ways to get around this--set up a "/etc/zoneinfo/defaultzone" file that contains the value to return, or generate a string from the abbreviations and other information stored in the "/etc/zoneinfo/localtime" file. But there doesn't seem to be any way to get unsetenv("TZ") (the Berkeley posting approach) to give the desired effect of getting your process--and any children that it later forks off--running on local wall clock time. If you take that approach on a System V system, and then fork off an "old" binary, the old binary will end up using GMT. I think the simple way out here just to avoid promising a way of getting your process and its unborn children running on local time. I still *do* want to provide a way to get your process--though not necessarily its unborn children-- running on local time; either unsetenv("TZ") or tzsetwall() does this job. The wise words from California are that unsetenv("TZ") is the way to go; are there any other words? --ado

1 0

Re: AT&T clock (synchronization)
by seismo!nbires!vianet!devine April 1, 1987

April 1, 1987

Mark writes: > I am unaware of any master clock in AT&T used for sync. I do know > that it's quite common for the clock in each central office to run > at a significantly different rate. I just got a reply back from a friend who works for AT&T in Denver. He says: > I'm not too sure about how central offices are set up for clock timing. > My past work in AT&T was with the long-distance calling switches (toll > offices). Back when AT&T was a monopoly, all the toll switching centers > kept their internal clocks in sync via a high-precision clock located > somewhere in Missouri. Presumably (but I'm not sure), the toll centers > passed along clock sync information to any connected tandem or local > central offices. With the breakup, all of this stuff must have been changed > dramatically. Bob Devine

1 0

Re: Addition to Arthur Olsen/4.3bsd table-driven ctime
by seismo!berkeley.edu!bostic%okeeffe March 31, 1987

March 31, 1987

> I cannot be sure that my customers will install the necessary > timezone support files. Therefore, any copy of ctime linked into > my programs must work "correctly" when transported to a machine > without /etc/zoneinfo in place. Neither can you be sure that they will install /etc/passwd. You don't handle that, why handle this? Not that I think that defaulting to GMT is correct, I don't. However, I can trivially fix it so that worst case is off by an hour. Which seems sufficient. > I insist that both timezone and DST corrections be applied, even in > the absence of /etc/zoneinfo, which means that ctime must carry > some DST information along in its data segment, so it can perform > at least as well as it used to. This is unclear to me. > The next question is, should any DST information hard-compiled > into ctime be fixed to reflect recent changes? The surprising > answer is "no." That means that your worst case is off by an hour. (Ignoring double DST.) The same as 4.3's. What am I missing? --keith > Steve Summit > stevesu(a)copper.tek.com

1 0