Hacker News Reader

Comment by gduffy | Hacker News Reader

Every rung you climb between client-server towards fully distributed increases software difficulty non-linearly.

It seems simple at the outset, but once you actually try to build a complex business/product like this one you realize you have to start with something simple just to get the money to fund something more complex/better.

RE: internet service loss leader model, I think it can be beat with a better product and a better model. But someone is still always going to need to pay to maintain and update software, and it seems fair to profit off of that as long as you allow for competition (& that's where I believe the law should e better protect consumers).

View on HN

> Every rung you climb between client-server towards fully distributed increases software difficulty non-linearly.

That's true. In my mind, fully distributed doesn't have to be the goal. I believe the number one problem for getting people to talk to their smart devices are NATs. I imagine a cloud service responsible only for NAT punching, and all the actual communication between user's smartphone and smart device happening directly (or rather, between the smartphone and home hub). It's probably more complex in practice than I think it is, but I can't think of an obvious show stopper.

> But someone is still always going to need to pay to maintain and update software

I think this is mostly a self-inflicted problem (or rather, a problem created and then used as an additional justification for subscription models). E.g. for a lightbulb, there's only few bytes of data that needs to be transmitted over the control channel. On/off state, color, intensity - setting them in one direction, reporting in another. That + overhead of whatever communication protocol is used. Such a device doesn't need an update. There's nothing to update there. The hub might, but arguably, hubs are designed overcomplicated too. But vendors seem to like to put a whole software stack on the devices, which now creates an attack surface that doesn't need to exist in the first place - and suddenly, security updates are required.

I think I understand where you are coming from, and I've thought similarly before.

But billion-dollar companies have been made almost purely on "NAT punching". I've written the code, and it's more complex than it seems.

And RE: your lightbulb example, I love it, because it I will now use it to illustrate how even seemingly simple devices require ongoing software maintenance. Is it using a wireless protocol compatible with other devices? Does it use encryption/authentication (e.g. to keep the neighborhood hacker kid from controlling my lights)? Does the setup process require interoperation with a changing set of personal devices (phones, etc)? All of these things could require software updates, see e.g. heartbleed. And if you have a software update system, it now needs maintenance as well...

Not to mention if consumers want their hardware investment to continue paying dividends through new software features. That part should definitely be opt-in and open to competition.

But those engineers you hire to maintain your software aren't commodities. They have shifting interests, bills to pay, and boot-up time to re-remember all of the old code. Costs will be lower to keep them employed and making continuous improvements once a product reaches sufficient scale.

Competition would prove out which model is best, though, so no need to think too hard about it, we just need to improve antitrust/competition law.

Thanks for the clarification on both points.

You've changed my mind a bit about the update capability - I suppose any wireless protocol necessitates a software update capability because it's exploitable remotely (e.g. from outside the house), and you'll never get it bug or vulnerability free the first time.

But this then calls into question the utility of consumer-level IoT as a whole. It's nice to be able to operate devices remotely from wherever you are, but this immediately creates a very large category of problems.

> Competition would prove out which model is best, though, so no need to think too hard about it.

Unfortunately, I'm not convinced of that, for several reasons. Information asymmetry - non-tech consumers can't evaluate these products, so vendors designing bad products have competitive advantage. Thanks to recurring revenue, service-backed devices can be at much lower price points than their service-independent counterparts, and most customers are very price-sensitive. Add in surveillance and data mining, and the price can be lowered even further. User-hostile business models have a distinct competitive advantage, because they offer immediate benefits but the costs are deferred. Therefore, I don't think competition alone is going to solve it.

Being capable of changing one's mind is a great thing. Thanks for the honest conversation – that's what's great about HN these days!

We competed head-on every single day with companies that lied like it was a national pastime, and we beat them handily. I think it is because good people tend to make the best products. There's an efficiency increase from passion, and all it takes is one good leader to unlock a team of hundreds or thousands of good people. I believe it is enough to take to the bank against shady practices.

That said, I do support better antitrust laws, we need to update them for the 21st century tech oligopolies + IP-stealing/currency-manipulating nation-states.

But all it takes is one good motivated person like you or me to Make The World A Better Place (™ HBO Silicon Valley). So let's get to it...!

> In my mind, fully distributed doesn't have to be the goal. I believe the number one problem for getting people to talk to their smart devices are NATs.

Ubiquiti does a good job with this - they provide the interface to get into my equipment remotely with little setup, but don't send the data to their equipment.

Reply on news.ycombinator.com