Collective time travel

[simont]

Last night was the first change of the clocks since I got my latest radio-synchronised alarm clock.

This clock, unlike its predecessor, is unable to walk and think at the same time. Or rather, it's unable to both display the time and synchronise itself with radio signals; so once every 24 hours it shuts its time display down while it listens carefully to the radio for up to 15 minutes before being satisfied it's got it right. And it has to do this only once a day, because shutting the time down is annoying and has to be minimised.

It turns out, in a staggering display of careful, attentive-to-detail design, that the time of day at which it chooses to do that is just before the radio signal adjusts when the clocks go forward or back. So my alarm clock is still wrong today, though it will presumably be right tomorrow. Or at least it had better be.

Good grief, who can have thought that up? It's not even as if you need to fully comprehend the MSF signal to know the clocks have changed: there's a one-bit DST flag broadcast every minute.

While I'm ranting, I'm also not fond of the way the clock change is officially mandated to happen. Instead of having certain times of day sometimes happen twice or not at all, it'd have been much better if they'd arranged that the hour between (say) 1am and 2am either went at double speed or at half speed, but remained monotonic. The only possible excuse is that the people who devised the current scheme had never heard of cron(1).

Comments

Not a pun

[bugshaw.livejournal.com]

I was going to put in a vote for missing an hour/having one twice rather than slowing down/speeding up time, as in computer logs it is much more obvious that Something Odd Has Happened when there is an hour with no events rather than two hours with half the usual number of events.

But what happens with logging in the hour that happens twice? If you have events at 01:05, 01:20, 01:59, then 01:04, 01:30, do they get sorted by time as if the 01:04 event happened first? Or does the timestamp include information about DST? Or is the order taken as the order in which log entries were appended, regardless of time stamp?

Re: Not a pun

[mooism.livejournal.com]

Ideally timestamps in logs are in GMT.

[mooism.livejournal.com]

Having the clock go at half/double speed would make keeping to a bus timetable somewhat challenging.

[mooism.livejournal.com]

Especially since some bus drivers drive at double speed already.

[simont]

Well, bus companies could publish an alternate timetable if need be. Or we could arrange for the hour in question to be a bit later at night when buses generally weren't running anyway.

[lionsphil.livejournal.com]

Or we could just ditch the ridiculous system and keep to UTC.

Also, why does cron(1) work on local time anyway?

[simont]

Well, you do want some kinds of cron job to be tied to local time, if the intention was to schedule them to run when people had gone home / gone to bed / whatever.

[pm215]

I'm reminded of the way that at(1) accepts 'teatime'. I don't know if this is preserving the memory of the departmental teabreak at Bell Labs or Berkeley...

[lionsphil.livejournal.com]

Oh, and I can possibly top that for dumb alarm clock design. Mine is a DAB radio, and is set to wake me via said radio. It's a Bose, so supposedly quite reputable for not being cheap and nasty.

A while back, the station it's set to, Capital Gold, did some merge-rename stuff and ended up as just Gold. The radio couldn't keep up with this, and lost the current station setting, which is understandable. However, when it then needed to awaken me, what did it decide was the correct course of action?

• Play some other radio station?
• Fall back to the buzzer?
• Sit there and silently display on its little LCD that there was no signal?

For a device whose primary purpose is the emission of sound at a given time, I found option three a rather unhelpful design decision.

[feanelwa.livejournal.com]

Mine (one of mine) is a daylight alarm clock, it slowly turns on a light for the right time. It is foolish for three reasons:
1) It doesn't actually wake you up, though there is a buzzer option, but using it is annoying since out of the array of buttons on it none of them says "STOP BUZZING"
2) The variable resistor bit and possibly some more circuitry is worn out, so not only does it not come on at the right rate any more, but it also doesn't uniformly do it wrong. Sometimes it only comes on half an hour before wake up time when I told it to do an hour. Once it came on and the automatic turn-off didn't work for two hours, so I ws lying there in limbo thinking it was 8am until 10am. I had terrible flu, so didn't think clearly enough to wonder whether it could be trusted as an indication of the time or not, but helpfully it didn't matter that I was then late.
3) It gains time. I thought all my other clocks (battery operated) were getting slower but then realised the computer and my mobile phone were then also getting slower, so the better explanation is that the clock is getting faster. This is annoying. If I wanted it ten minutes ahead all the time I would set it ten minutes ahead all the time. And, if you reset the time to what it should be, within a few days it is about ten minutes fast again.

Mind you my computer isn't much better, since it still thinks I live in California.

[teleute.livejournal.com]

Mine is a little bedside alarm clock. It does the alarm bit just fine, but like yours gains time. I reckon mine is gaining about a minute every 2 days. It has done this since we bought it (although it took us a long time to work it out since at the time it was the only clock in the bedroom). I'm astounded that people can still make and sell clocks that don't actually function as clocks.

[deborah_c]

My ex's clock radio has a variation on this. It's got RDS, and will happily track time from it. Of course, this works only when the radio is on, so when the clocks change, it notices only after waking you up at the wrong time, unless you happen to turn the radio on after 2am, of course.

[cartesiandaemon.livejournal.com]

Hm, double speeding through an hour is a great idea, I don't know why it never occurred to me.

(I don't think it is more practical though. I think most normal people will set the time on an hour whilst they sleep, and not have to cope with it. Whereas trying to do anything while watching a clock whizz by would be fun but very disconcerting; and unhelpful if you try to specify any fraction of an hour.

And electronic devices probably have it easier adjusting an hour once. And anyone doing anything time-critical will have to deal with it.)

On the other hand, how about spreading the change over a day or two? That's even more mad, but possibly less perceptible for most normal operation...

[simont]

I think most normal people will set the time on an hour whilst they sleep, and not have to cope with it.

Well, normal people can still set their clocks forward before they go to bed or after they get up, or cope with the time around the clock change however else they want as long as they get in to work on time. The important thing is that some systems really need to have the officially right time, and some systems need to have a monotonic clock for cron jobs etc, and some systems need both, and the latter are ill served by the current system.

On the other hand, how about spreading the change over a day or two?

If I were designing a DST system (and didn't just decide to say "sod it" and do without) I would actually go even further than that: I'd arrange continuous slow adjustment over the course of the whole year, to keep sunrise within a reasonable period. No sudden changes ever.

Downside is that everybody's clocks would have to be much more intelligent to cope, but then if I were designing a DST system at all it could only be because I'd revolutionised the world with mad new SF technology and then had to take over running it myself because nobody else could understand my crazy hacks, so everyone would have mad SF clocks conveniently available anyway :-)

[cartesiandaemon.livejournal.com]

The important thing is that some systems really need to have the officially right time,

Yeah, ok, good point. It *does* do that. Very well. I just have a bad feeling. Say, trains have to have a fixed time. But I'm sure having "12:05 BST, 13:05 BST, 13:05 GMT, 14:05 GMT" is easier to cope with than "12:05, 13:02, 13:32, 14:05". Admittedly that's mainly for human interaction.

Downside is that everybody's clocks would have to be much more intelligent to cope,

But on the plus side, that's probably fairly easy to do now, and does mean everyone has to make damn sure they do it when they design a clock, rather than just hoping people work out the time change "somehow"

The perfect DST system

[fanf]

Your comment had me bursting with suppressed amusement :-)
http://catless.ncl.ac.uk/Risks/25.10.html#subj1
http://fanf.livejournal.com/85472.html

[fanf]

My MSF clock is also still showing GMT. I told it to resync yesterday, and I have done so again today. It's not implausible that NPL or their subcontractors failed to adjust the DST flag...

[sparrowsion]

My radio-synchronised clock, which is the indoor component of the weather station, chose About Now to run low on batteries, so that although the display was working and it could hear the outdoor (well, in-shed) sensor, it could no longer pick up the time signal. Which in this case is coming from Frankfurt.

The RDS-synced clock on the #2 stereo isn't really synced—you need to explicitly tell it to go and get the current time. And it's not actually that useful, as there isn't any mode in which the time is permanently displayed.

[bjh21.livejournal.com]

Hmm. I didn't use to entirely trust my MSF-synchronised alarm clock to correctly track the BST change, especially since it usually only fires up its radio once a day, at midnight. Nonetheless, it's got it right every time so far, and even woke me up at the right time on Sunday, which makes it cleverer than damerell.

[damerell.livejournal.com]

:-P

I only score two DNS - alarm clock so far. But bloody Stevenage is my nemesis, it seems.