• fubarx
    link
    fedilink
    English
    arrow-up
    5
    ·
    7 months ago

    I’ve had the pleasure of reading through the whole Matter spec for work (twice) and working on this stuff. Here are some issues:

    • It’s a very complicated spec. From onboarding a device to changing a value (on/off, dim-level), a developer has to take a LOT of steps. It takes a long time to develop code that gets things right.

    • The CSA (group that owns the Matter spec) has published a lot of sample and SDK code to hide the complexity and help people not have to reinvent the wheel. But that code sometimes has conflicts with the version actually running on the phones and home gateways out tbere. That makes testing hard.

    • The spec has a pretty complicated flow for adding a device to a home, but sort of punts on how the whole network has to coordinate with the cloud (if at all). Same with how to remotely control a Matter network when you’re away from home. So we’re back to needing custom apps from vendors.

    • Speaking of the network, device vendors LOVE hoovering usage data and sending it all to the cloud. Every time you flip that light switch or change the color of a bulb, it’s saved somewhere. Why? Because that’s what everyone else does. Matter allows you to have a local network that doesn’t need to talk to the cloud. In practice, you’re likely to use Apple’s Homekit or Google Home Automation to pair and operate the devices, so if you use one of those clients, only Google and Apple get that usage data. A lot of vendors are wondering why they should go to all this trouble and NOT get any data out of it.

    • To get vendors to adopt it, Matter has an anti-counterfeiting system baked into it that directly benefits manufacturers. But it requires setting up a complicated certification process and getting each product ‘approved’ by CSA. This adds lag time and has created a whole expensive ecosystem for device certification vendors. Add that as an ongoing cost for producing a Matter device.

    • That little sticker with a custom QRcode used to easily pair a device? It has to be generated and affixed at the factory. A lot of electronic factories aren’t set up to generate and print them, let alone stick them to the device. To be safe, you also need a backup copy of sticker in the user manual. That means user manuals need to be paired with each device. Guess what? Lots of factory lines aren’t set up for that either. More delays.

    • There’s more, but the biggest problem, IMO, is that Matter has a fixed device taxonomy. This means there is a defined device type of say, light bulb, and it has a predefined set of attributes and operations.

    If you want to innovate and add a crazy new feature that sets your product apart, you have to spend a lot of extra engineering effort to create your own custom attributes which may not be supported by your HomeKit or GHA apps. So you’re back to creating your own custom app or waiting for the next Matter spec to catch up. Fixed taxonomies are driven by central authorities and slow down the rate of progress. It’s baked into the Matter spec.

    The upshot is, it’s a good idea having a standard. In practice, it’s adding cost and delays to every product. It will take a long time before vendors have it all streamlined and can crank out new products with ease. We’ll all have to wait until that happens.

    Thanks for coming to my TED talk.