Oh really, and how does the system know what you did to the lights while the internet was down?
I understand why it seems like this would be something easy. That’s the kind of world we live in since broadband Streaming was invented. But in fact, this is very hard to do, and very expensive, if you want to have $25 sensors and $15 lightbulbs. And I understand why it seems really frustrating that it isn’t there.
And exactly what you described is available now with Wi-Fi devices. But those devices cost $125 per light bulb.
TOT: The Twitter of Things
The whole point of mesh like zigbee is that in fact The hub doesn’t know the state of every device every second. Things just check in periodically. If someone turns a lightbulb off at the switch, or if the power to the network goes off, or (the real rocket science for which zigbee was invented) someone takes the batteries out of asensor to change them, The hub doesn’t know.
I know that seems weird. But it’s not stupid. Be patient with me for a second, because It’s actually pretty brilliant engineering. Because the lightbulb and hub are not in constant contact, but just checking in from time to time, the lightbulb can be a really stupid device. (The device, indeed, is stupid. The engineers who invented it are not.)
It has absolutely the minimum amount of hardware it needs to operate its radio signal to send and receive messages. And those messages are super tiny. It’s kind of like Twitter for devices. Except instead of over 100 characters all the messages there are maybe 15. And there are no Emoji’s. 
The messages are things like: 17, on. Meaning device 17 should turn on. And the hub doesn’t even send the message directly. It’s just handed to the closest guy who keeps passing it along until it gets to 17. Who just turns on.
If you want the lightbulb to remember anything, including the last state it was in, it would have to have a much more complicated computer inside of it. It would have to be able to store information, and it would have to have be able to store it even when for three days when it had no power at all. That, again, is rocket science.
And it’s not just that the computer inside the lightbulb would have to be smarter. And would have to have its own energy source. And would be physically bigger. It would run hotter. It would draw a lot of energy. At least a lot compared to the very little power that Zigbee bulbs use now.
There’s nothing wrong with having a lightbulb which is smart enough to remember what it was doing last time all the power was cut to it. But it’s going to cost a lot more than $15 to buy.
Too frustrating for a mass market?
Apple believes that most people are going to have exactly the same reaction that you’re having. That it’s stupid that each individual device doesn’t know everything that happened to it yesterday, and doesn’t have a whole set of rules stored for what it should do today, and that if people don’t get devices that all act basically like they are as smart as iPhones then The customers are going to say that the engineers are idiots.
So, in the HomeKit universe, devices will meet the standards of people who are used to Wi-Fi devices. But the door sensor is going to cost $125. And you’re going to have a much more limited set of devices to choose from.
Again, I understand your frustration. And if you hate the way Zigbee bulbs work after a power outage, just don’t buy them. They’re not the right tool for you. Buy smart switches instead, and use them with dumb LEDs.
But an engineering problem is rarely just “make the device do this.” It’s “make a device do this, for this price, and have the power batteries last for this long.” And once you throw cost in dollars and energy into a problem, some apparently simple problems become really really hard. A networked $15 lightbulb is one of them.
Not all problems have the same solutions
Zigbee solves a really hard problem really well. It’s just not the problem that many residential home automation customers care about. Zigbee is a gloriously beautiful bicycle that is comfortable to ride and easy to fold up and carry on the bus. And most people want a car that can take two kids and a dog to the park. Or a bunch of friends to A restaurant. Or at least a couple bags of groceries home from the store. 
But as soon as we start talking about raincovers and big baskets and adding an engine so you don’t have to pedal uphill, we’re solving a different problem. And the bicycle doesn’t function as well as a bicycle anymore. Let alone one that folds up and fits in a totebag.
It’s not that the Zigbee folding bicycle is missing pieces that somebody should’ve been smart enough to see it needed. It’s that it was designed to solve a different problem.
If it’s not solving the problem that you have, you get a car. 
Submitted with respect.
4 Likes
You cannot make changes to the system without access to the cloud. When access to the cloud comes back you put the lights in the last known state.
Things like power outages and internet outages are not discrete events that can be treated as such. They are by definition non-deterministic events, where the relationship of the system state and the real world state are no longer bound together in any way. But you are proposing to ignore that, and pretend that the outage was a discrete event in which the system state has remained pure and the desired state for the system post outage. That might be true, or it might not be true, and how is the system to know which? An outage may have created circumstances in the real world system for which automatically “restoring” a previous state is the wrong thing to do. Light switches still function when the internet is out. Time still passes. The system state can’t possibly be in any other condition relative to the real world state than unknown at the moment power is restored.
2 Likes
Osram has implemented this option via their bulbs and gateway, replace all your GE Links and be merry.
It wasn’t that the Zigbee Org couldn’t figure it out, but that they made a choice that the best way to gain adoption of these smart bulbs was to make sure they also worked like dumb bulbs when using a switch. It was for many practical reasons including safety that this option was preferred. Power outages are relatively rare occurrences. An uninformed user (maybe a spouse 
) flipping a switch and getting no response from the bulb is much more likely and detrimental to the adoption of these.
2 Likes
This is not correct. Most Zigbee devices run on 32bit Cortex M0 processor that have plenty of performance. They also have EEPROM memory that can store values for an unlimited amount of time with zero power.
I am just shocked that Zigbee has been around for so long and is so broke. Its number one function was home automation and the number one profile was lighting. It fails at that function.
One question? If nothing can change with out access to the cloud how would one make changes to the system when there is no access to the cloud?
The hub is the Zigbee coordinator. Without the coordinator none of the states can change.
The smartthings hub includes the zigbee Coordinator, but the account authentication is based on cloud architecture.
Staples connect allows you to continue to access their system when your Internet is down using your phone on your local network.
Smartthings does not. If the hub doesn’t have an Internet connection, your phone app won’t talk to it. As a customer, we were not given a direct interface to the Zigbee Coordinator or the Z wave controller. We have to go through our account.
So even with the new hub, when your Internet is down, whatever was already locally stored on your hub can run, but you cannot make any changes that require touching your account records.
They’ve already told people to preprogram mode changes and arm/disarm changes into a minimote or they won’t be able to change them if the Internet connection is lost.
That’s just the service that they’re selling. And that one seems weird design to me!
Light switches still work, and change the state of things. Other controllers may still work. Time passes (was light, now dark). People were home, but have now left. Doors were closed, now open. Etc. The world in which the system lives is not just zigbee lights. But, irrespective, you want to put it into a state that existed some time ago, and you are making assumptions about that being the preferred state. The system state may not even be correct, if the outage occurred mid-processing of some event. How do you know that everything is so discrete? How do you know where in the world of asynchronous events the system was at the moment of outage? Or at the moment of restoration? Only by assuming something that may or may not be correct.
But, moot point. The likelihood that ST would make this change, even given your insistence as to its correctness, is zero.
2 Likes
In my option you are never going to make changes during a power outage because you do not have power. I think restoring the system to the last know state is the best solution.
1 Like
I live in a semirural area and we have power outages several times a year. First thing we do in a power outage is go around the house and turn a bunch of switches off. While the power is off. That’s been SOP since my childhood, long before I used a wheelchair. I think that’s pretty typical.
If the power went off at 8 PM while people watching television in the living room, and then everybody went to bed, I definitely don’t want the house restored to where it was when the power comes back on it 2 a.m. Which was your basic point about the lights in the first place.
My ideal experience from the user side would be that the device controller, not the devices themselves, could distinguish between a lost power condition and a power off instruction, and offer a rules engine that each person individually set for each device for what to do after a lost power condition.
I actually do have that for some (very expensive) medical equipment.
It’s the old elevator problem. Power goes out. Elevator stops. When the power comes back on, do you want the elevator to immediately resume moving? What if people had forced open the doors and were starting to climb out?
There’s no one safe answer to the “resume previous state?” Question. Any of my lift devices, whether it’s a bed hoist or a bath lift or an elevator, or a garage door, or an automatic window closer should NOT reuse previous state after a power loss. But the aquarium pump and the furnace can. At least at my house.
1 Like
That is great that OSRAM implemented this function. It should have been in the spec 15 years ago. I understand the safety aspect. With the OSRAM system you can put some bulbs in a mode were they come on at full brightness out of an outage. But decorative bulbs can act different.
1 Like
We have maybe 5 outages a year. Most are sub 3 minutes. I have a full house generator but when the power goes out I have zero power until it switches over.
Well, you’re obviously convinced of this approach irrespective of what anyone else says. Nice discussion.
I prefer the present approach to power restoration: process new events as they occur, make no assumptions, do nothing in response to that non-event.
I do use the Power Notify app, tied to a Gen 1 SmartThings motion sensor. That app turns on my landscape lights if it’s after sunset when power is restored. If the outage occurs at 2:00 AM, that was the wrong thing to do. If the power went out at midnight, the landscape lights were on. So, simply restoring the state wouldn’t be the right choice, and even my automatically turning them on after sunset isn’t the right choice. The point is, there is no perfect choice unless one were to implement some humongous “put the system into the right state” app that ran through all of the state logic of the entire system when power is restored, and that app could somehow determine by inspection all of the state changes that should have happened during the outage that didn’t happen, but now should happen. Can of worms!
2 Likes
This is what happen to me.
We went to Florida for 9 days. The power went out day 2. All my 9 outside lights came out a full brightness. 4 lights in the master bedroom came on at full brightness. 3 lights over my mantel came on a full brightness. 4 lights above the kitchen table came on a full brightness. 1 light in my office came on a full brightness and 1 light in my wife’s office came on a full brightness.
The entire house stayed in this state for 7 days except the outside lights turn off at sunrise the 3rd day. I am sure my house was glowing that week.
Here’s another example using only zigbee lights. Let’s assume the starting condition is that two zigbee lamps in the baby’s room are off when the power goes out. Some time passes, and now it’s time to change baby’s diaper and get pajamas on. Power comes back on, but internet does not for a few more minutes. Those zigbee lights turn on. Great, Mom needs the lights to change baby. Now the internet comes back, and with your solution, ST will go turn those lights off. Mom screams from dark room with poopy diaper in hand. Not good. At that point there will be lots of angry posts about dang ST turning off my lights after power restored.
2 Likes
I understand. It’s very frustrating.
Because I have housemates and our schedules are very irregular, when we’re home, I can’t just set something to automatically turn off all lights at midnight.
If we were all away on vacation, I wouldn’t have a problem having a vacation mode routine that turned all the lights off at one in the morning every night.
And if I got a power loss and restore notification while I knew everyone was out if the house, I would just manually toggle off all the lights, and manually toggle back on whatever I wanted to back on, and then let the rest of my automatic stuff run. If somebody was still home, I would let them deal with it. 
But that’s just me. We don’t have that many power outages. It’s an unusual situation. I’m fine with the cheap and dirty solution of having the human make a couple of adjustments.
Someone whose vacation was going to be 10 days hiking the Appalachian Trail might rather just set their vacation routine to automatically turn lights off at least once a day, even if it was in a more random pattern.
It sounds to me like the biggest issue for you may have been the surprise of discovering that this was the expected behavior after a power loss. It is in the user guides for all these devices, but I know it can still feel unexpected.
The problem is that the LED bulbs may not respond after an outage. That said… better than do nothing I suppose.
1 Like
I only see this with the GE links, which is reportedly a known firmware problem. (I’m replacing all of mine with the new hue whites.)
Have you seen it with any other brands?
I presume you’re being hyperbolic.
While Apple users are accustomed to paying a premium, there is an upper limit that the market to bear for something as basic and required in mass as door/window contact sensors.
I think @Judgless makes many good points. The benefits of cheap and low power ZigBee have come with disadvantages. That doesn’t mean that the compromises made should be considered acceptable. It also doesn’t mean that the system couldn’t do better without requiring a completely different architecture.
“Choice is good” – @JDRoberts has said.
Having at least the choice of feasible behaviors for light bulbs after a power failure is valuable.
1 Like