My post on Lowe’s killing off their home IoT product, Iris, kicked off an expert-level social media discussion on how IoT systems “ought” to be. Here’s a summary of what people said, grouped by subject, plus some implications. I figure it might be useful to keep this around for myself, and others, so I am posting this.

This is all about home IoT.

On communications and protocols:

• All protocols need to be fully documented. No secret commands anywhere.
• Use standards for communications on all levels.
• Only make devices accessible over the internet that are proactively maintained (for security reasons)
• Do not require any device to have a data connection to the internet, ever. In particular, devices must be functional without depending on a vendor cloud.

On identification and security:

• Every device must have a globally unique serial number, which is printed on the device.
• Every device must have a unique public key, which is printed on the device.
• Devices must authenticate to each other when interacting with each other.

On who has control:

• The customer must be able to continue using IoT products even after the vendor goes under or discontinues the business.
• The central control logic needs to be hackable and replaceable by the customer, so it should be open source or replaceable with open source.

On cost vs benefit:

• Products must not cost more than the customer is willing to write off if the vendor goes under, otherwise nobody will buy them.
• Maintenance costs for many products are far too high compared to their benefits.
• We need more valuable solutions to actual problems than just gimmicks like turning the lights green from the mobile phone.

On configuration and maintenance:

• Make it easy to reset devices to factory settings.
• Must not require complex scene setup (too little benefit for the effort)
• Must become much simpler to install and configure (e.g. so tradespeople can offer installation contracts that are affordable for not just to the very high end of the market)

On customizability and integratability:

• Making components from multiple vendor work together must become much simpler (e.g. cameras, open-source software)
• Devices need to be able to talk to each other directly, and control each other directly. (A mobile app is insufficient)
• The whole system needs to be programmable.
• Components need to be designed to be integratable into existing systems.

On home networks:

• Home networks need to gain features for observability and controllability that today only exist for the enterprise market (e.g. how do you identify a misfunctioning device on your network?)

On the distribution / value-added channel:

• Need simple, cost-effective solutions to home network and configuration problems that today require the expertise of an expensive networking professional with substantial qualifications.
• Need features that allow tradespeople to do their jobs at least as well as with traditional solutions (e.g. the Next thermostat has no mode to allow a HVAC engineer to debug the furnace)
• We haven’t figured out the channel yet for the distribution of IoT products: home improvement retailers do not have the expertise.

On safety:

• Systems have to become safe to use even “when a complete moron uses it”.

There is probably more, but this is an excellent collection! It also very much aligns with the recent discussion between Peter Hoddie and myself. Who’s up for collaborating to make it so?

Credits and thanks to Mark Atwood, Bob Frankston, Kathy Giori, Phillip Hallam-Baker, Dan Lyke and Jeff Nolan.