I use apps that interact with their specific devices. An example is Roku. The device is Roku and I use their Android App. Another is my Denon Audio receiver and it too has an app to control it. I'm looking to write app/hardware service. how does an app detect device on network and vice-versa?
Disclaimer: I don't know what Roku is but what you are describing sounds like service discovery as part of zeroconf.
One relatively known example is DNS-SD. You can think of it as sending periodically multicast UDP messages containing stuff such as "I am at address X and I provide service Y" to some well-known address. Whoever is interested will pick up the information associated.
Related
When connection to a custom BLE device from an app (android and ios) I have two options
a) first pair the device on the phone (like a headset or other devices) and then use it from the app to read and write data
b) without pairing on the phone, go inside the app, search for a device with a specific name or address, and just connect to it and read and write data.
when is one preferred over the other one ?
Its just when it needs to interact in the background or am I missing something?
The devices you see in your phones Bluetooth settings are Bluetooth Classic devices. There are some devices using both BLE and Bluetooth Classic which show up in the settings too, but the settings page is only for the Bluetooth Classic part of the device. Headphones for example are streaming music trough the Bluetooth Classic profile A2DP but might offer BLE services for additional features.
You always have to use a special app to communicate with the BLE part of a device. This could be an app provided by a manufacturer or a generic BLE scanner app such as nRF Connect.
As alexander.cpp already told you in his answer pairing (the exchange of keys) and bonding (saving of said keys) is not required for the communication with BLE devices and only needed if the device requires a secured connection.
Bonding (technically correct term for BLE, often called Pairing in non-technical speech) is mostly for security, we can send sensitive information because the receiver is verified and data is transferred in encrypted form.
Prefer to use bonding if the data is sensitive (means users feel uncomfortable if somebody they don't know receive their data). For example, I consider "current temperature = 25" as NOT sensitive. For detailed explanation, find "Bonding with a BLE device" in BLE guide by PunchThrough.
Also this is a good answer to a similar question: https://stackoverflow.com/a/42916081/10380092.
Your second question about background - no, bonding is not related to interaction in background.
I am looking for a way to scan a device I own and discover "what it can do".
In other words, I'd like to know if a device is able to describe the way you have to communicate with it in order to build some application around it.
In my case it is a simple Christmas light that I'd like to play with but this could be used in different situations.
For classic bluetooth (BR/EDR):
When scanning for bluetooth devices (Inquiry), the bluetooth device will send an inquiry response (if it wants to be discovered) and maybe also an extended inquiry response (EIR). This EIR may already contain a list of services, the devices supports. This is a very fast way to get a picture of a remote device.
Moreover, the service discovery protocol (SDP) gives more information on a device. This takes some more steps. In SDP two devices can exchange their capabilites in kind of ping pong process.
For BLE:
After connecting a BLE device usually a service discovery takes place. The BLE peripheral (e.g. headset or a light) reports its capabilities to the central (e.g. smart phone). Some of these services have predefines functionalities. Additionally, it is free to the manufacturer to add custom services.
Chromecast v2 devices announce their presence using mDNS, and they are discoverable when performing mDNS queries for _googlecast._tcp.
DNS-SD / Bonjour also support the concept of "wide-area discovery", which makes use of standard unicast DNS queries to find devices. This can be useful for more complex networking scenarios - i.e your Chromecast devices may be in one VLAN but your sender devices are in another.
However, when trawling Chromium source I found this code in mdns_api.cc which seems to indicate that the Chrome extension will only search for _googlecast._tcp.local - completely preventing wide-area DNS discovery of other Chromecast devices. Based on anecdotal testing, it seems the Chromecast iOS app and SDK also have this behaviour.
Why does the official discovery mechanism for the Chromecast Chrome extension explicitly only discover link-local Chromecast devices?
They probably didn't do "wide-area discovery" because there are a lot of gotchas in implementing it and there isn't much gain to be had.
Let's suppose that they did implement it (from looking at the spec, it doesn't seem to be technically difficult; seems like it is mostly an edge-case problem).
You'd need:
A domain under your control. Could be a local-only domain.
Not many people do this.
Multiple VLANs that you want to use.
Most homes only have a single VLAN.
A DNS-SD server that supports Wide-Area Discovery.
This is probably the easiest thing to have. Even then, most people wouldn't do it.
A UX flow to input the various WAD servers that you want to ping for devices.
This is the hard part for Google since it needs to be consistent across all SDKs and using a WAD-discovered Chromecast would cause all local-media-server Apps to not work (e.g. Plex).
(1) - (3) are why Google wouldn't make this a priority. Chromecast is, after all, a consumer device and consumers tend to have simple networking situations. (4) is why it isn't low-hanging fruit.
You'd also need to solve a few problems:
What happens when you are connected to a Chromecast on a separate VLAN and then remove it's WAD server?
What if the WAD server goes offline?
What happens when you have multiple Chromecasts with the same name?
What if they also have the same IP address (possible since VLANs can have overlapping IP spaces)?
What happens when a Chromecast is discoverable using WAD but not reachable?
Should they attempt a connection to every single Chromecast you discover to test that it is reachable?
How would you scale this out to work with 100s of Chromecasts in VLANs that could be "far" away?
It is these problems that I think would prevent Google from implementing WAD even if they wanted to.
For example, if I have a friend over and he wants to show me a video using a given app that runs both on my device and his device. Could that app display a QR code on the screen or something that he could scan and instantly be granted access to my Chromecast device?
As Ali mentioned, Chromecast devices are discovered and apps launched via local network applications. One an app is started, it could easily connect with a cloud service that allows other (non-local) devices to talk w/ your Chromecast via the cloud service. A Chromecast Receiver application is just a HTML5 application (HTML, CSS, and JavaScript). You can really do what ever you want once your application gets launched.
If displaying a QR code that allows some kind of rendezvous with your cloud application is what you want to do, you can certainly do that.
I presume your friend's mobile device is on the same wifi network, right? Currently, a chromecast device has no identity outside of its local wifi network, so if the sender is not on the same network as the receiver, there is no way they can exchange messages. Back to your question, if your friend is on your network, then he could see your device except from those applications for which your device is not whitelisted for. Is that the case you want to handle through, say, a QR code? If so, that is currently not doable either since whitelisting is not just a local setup. Maybe I misunderstood your question?
Based on your questions, you are saying that both you and your friend have the same app. If so, and if your friend connects to your wifi network, then he will see your chromecast (you do not whitelist a device for a phone, you whitelist a device for an app id and as long as your friend has the same app (hence the same app id I resume), your whitelisted device will be discoverable by his phone. On the other hand, if you do not want to give him credentials to get on your network, then you need a cloud backend and a lot of work, since although your chromecast device can send a message to cloud and your cloud service can notify the other user's phone (using, say notification or some other mechanism that you employ in your app), the reverse (i.e. sending a message from your friend's phone to your chromecast (through your cloud service)) is much harder. Your friend's phone can send your phone a message (again via a back end service, a bluetooth communication, NFC, etc and then your phone using your app can send that to the chromecast receiver but I am sure you are getting the idea that it is a lot of work. Signing up on your wifi network can be made easier with a QR code or something so at this pint, that would be the easiest solution.
The Chromecast device is a "receiver device [that] runs a scaled-down Chrome browser with a receiver application". Can I download and install this receiver app on a chrome browser for example on my Windows notebook?
I have implemented a complete chromecast v2 receiver, called YouMap ChromeCast Receiver, available in Google play store and Amazon store, xda-developer thread here: http://forum.xda-developers.com/android-tv/chromecast/app-youmap-chromecast-receiver-android-t3161851
The current Chromecast protocol is a completely different one from the original DIAL based protocol. Right now, only YouTube still uses the old protocol, which chromecast maintains its backward compatibility.
The discovery is mDNS, exactly same as Apple TV bonjour protocol.
The most difficult part is device authentication, the sender and the receiver perform handshakes by exchanging keys and certificates in a way extremely difficult to crack. AppleTV does the same using FairPlay encryption.
The next difficult part is the mirroring protocol, which is also very complicated, need to deal with packet splits, packet retransmissions. Overall, chromecast mirroring protocol is well designed, better than miracast, better than AirPlay mirroring (I have also implemented both of them, so I know what I am talking about).
When I get chances, will write more here.
The chromecast device works using the DIAL protocol. It is completely possible to emulate this protocol using some simple code to listen on the multicast group for discovery and then handle the HTTP requests to launch applications. It is then the launched application that communicates with the casting device, I believe using the RAMP protocol.
Luckily for us the applications that the chromecast device uses are mostly web applications meaning our device emulator just needs to launch a web browser and point it to a specific url when it receives an application request.
For example the youtube app, after device discovery and establishing where the applications are located (part of DIAL). Will send a HTTP POST request containing a pairing key to /<apps url>/YouTube. All the emulating device needs to do now is open https://www.youtube.com/tv?<pairing key> in a browser window. From here, I believe, communication for controlling the youtube app is not sent through the casting device but through the open tabs on the casting device and the emulator.
This is my understanding of how the chromecast device works and specifically the youtube app from looking at https://github.com/dz0ny/leapcast which is a python emulator that has youtube and google music working.
Google is in progress of open sourcing some part of the chrome cast.
https://code.google.com/p/chromium/codesearch#chromium/src/chromecast/
https://code.google.com/p/chromium/issues/list?q=label:Chromecast
So theoretically you can build a similar device.