Thread by @swardley

X : How do you define serverless?
Me : Roughly speaking it's an event driven, utility based, stateless, code execution environment in which you write code and consume services. A boundary condition is "write code" i.e. any lower than this and it's not serverless.

X : What about containers?
Me : Part of the infrastructure world, but a higher abstraction than servers. This might be used as a component to build a serverless environment but if you're operating at a serverless level then you don't care whether it's on a chip or whatever.

X : How do you mean you don't care. Containers are an important part of serverless.
Me : No. If you're diving into containers, you're diving out of the serverless world. The boundary condition is "write code" and above. Anything below that is taking you out of this world.

X : What if you're build your own serverless environment?
Me : Then you're stepping out of the serverless world and diving into infrastructure. Whether you want to do this, well that's upto you. I can't think of many valid cases for doing this. Certainly not sustainable ones.

... think of it like this, if you're in the hotel business then you don't necessarily want to build a farm just because your hotels have room service. Of course, farms are part of the value chain but so are water companies. You don't want to build one of those either.

... you could but it's an odd idea. We normally focus on the level we're at and others will build on top i.e. travel companies using hotel chains. You could own the entire value chain but where do you stop? Pharma for veterinary services for cows for steak in our room service?

So, you might say ... "let's build our own serverless" but unless "selling serverless" is your business then that's inertia mixed with some vendor selling a story based on sunk capital because you have to ask, why stop there? Why not our own chip plant, our own power etc etc.

X : Serverless contains servers.
Me : Hmmm, I don't think you understand how "Thought Leadership" tends to "evolve" in IT. It's not a pretty business but I'll try and explain in a single graphic.

X : I disagree with your definition.
Me : That's ok, it's not a crime.
X : I think there are cases for building your own.
Me : That is a crime.
X : The cases are valid.
Me : It's your thinking I have concerns with. As Thought Lord I order you to avoid attempting it again.

X : &*^! you
Me : Look, I'll keep it simple. The fundamental disagreement you have with the definition is likely to stem from the fact that you want to build your own. But building your own is not part of serverless. Make a choice ... head towards serverless or not.

X : Building your own serverless environment is still part of serverless.
Me : Not, it's not. Orchestration, Kubernetes, Containers, Servers, Data Centres and so on are not serverless but they maybe part of the serverless value chain. They're the infrastructure bit of it.

X : But look at electricity, it's decentralising with micro generation.
Me : The forces that cause a yo-yo between centralisation vs decentralisation in a stable utility market are different from the forces that cause industrialisation and whether the initial market centralises.

i.e.
a) industrialisation (evolution to a commodity) is driven by supply and demand competition.
b) the Yo-Yo tends to be driven by constraints in underlying components.
c) the initial centralisation (or not) by the strategic play of early competitors.

In the case of serverless, a shift of framework from product (LAMP et al) to utility (evolution to an industrialised form) is driven by supply & demand. Market is not a stable utility market (no Yo-Yo) and initial centralisation is because Amazon plays well, competitors don't.

... at some future point (e.g. 20 to 30 years down the road) we might see a bout of decentralisation in both compute and serverless with alternative providers once the market has become a stable utility i.e. many Lambda providers. But not now, the forces necessary aren't there.

If you wanted to see a decentralised competitive market then to be frank, competitors should have got off their bottoms in serverless well before Lambda launched. It's not like this wasn't predictable. There was a chance around cloud foundry but that got sucked into containers.