[openthc]

hahaa; wow. So, we were doing this sensor project; and I picked "boring" things like raspberry-pi, python scripts, wiring the GPIO with a screw-down terminal or using SDR for other off-the shelf sensors to build the network. And in one of the demos showing our very low-budget type project a reviewer said: "you should look at IoT managers like the G offering, we're using it". So, they declined to use our methods and built their own around G-IoT. It's important to "own" what you can in your stack otherwise this vendor-driven-churn is forced on you and is outside a schedule you control. Sure, 365 days is a lot of time to migrate -- which, IME, leads to "we can fix this later" which then leads to "oh crap!"

[Nextgrid]

I hate the current trend of smearing your business logic across dozens of third-party providers with various business goals that rarely align with yours.

I'm also not confident that the constant churn of keeping up with API changes (if not outright deprecations like this) and costs of the third-party services end up costing you more than just doing it yourself.

Finally, what we've seen with Okta, Twilio, most recently MailChimp (which was used to attempt to attack DigitalOcean customers among others) clearly shows that these companies aren't magic and may not actually be any better than doing it yourself when it comes to security.

[c_o_n_v_e_x]

>I hate the current trend of smearing your business logic across dozens of third-party providers with various business goals that rarely align with yours. I'm also not confident that the constant churn of keeping up with API changes (if not outright deprecations like this) and costs of the third-party services end up costing you more than just doing it yourself.

I suppose the problem is that it's hard to evalute the total cost of ownership. Migration to a 3rd party and operations costs can be reasonably forecasted, but how do you evaluate business or product continuity risk? Without being able to evaluate that risk, the 3rd party service always seems artifically "cheap."

Can you buy insurance against vendors dropping their services? I've licensed some tech before that included statements that if the company were to go out of business, the code would be open sourced so as to not leave people high and dry.

[tommiegannert]

> Okta, Twilio, most recently MailChimp

One issue is that the bigger you become, the more interesting you are as a target to break. So if you're a small company using a big company as a vendor, you've probably introduced a weaker link.

But if you're still small as a SaaS, trust is an issue to acquiring customers. I'm wondering where the sweetspot is.

[andrewxdiamond]

I think it’s important to “own” what makes you different.

Running your own servers isn’t that important and probably isn’t what makes you different. EC2, Azure VMs or whatever short term project Google is running for compute are all extremely comparable, and you loose very little by using them.

But if your business is owning and operating an IoT platform for your customers, you should invest in a high quality solution, not just buying the off-the-shelf tool. It may be that a provider’s offering is better than what you can do, but it better be a lot better if you use it.

The point is to invest heavily in what actually makes a difference, vertically integrate what matters, externalize what doesn’t.

[robertlagrant]

Slightly disagree. The important thing is to externalise things that multiple providers support. So externalise hosting Postgres, because 20 companies exist that do that. Don't externalise a one of a kind service you can't build against using open standards. Insource that.

[drunkenmagician]

A slight variation I prefer, and strongly advocate is ‘own (or control) what makes you successful’. It may not be unique, and doesn’t imply always building it yourself, but you should be in control of those things.

[gerdesj]

I think it is important to mostly be in charge of your own destiny too.

When my doorbell (Doorbird) goes off at home the following happens:

* A real chime sounds (battery backed, mains powered, no internet required) * Native app goes ape shit and twitters (often about two hours later) * Home Assistant app kicks in on queue and does what I tell it to: Speakers speak, SMS sent etc