Avoid complex infrastructure when building simple things
Julian Friedman
816

You’re not right, but you’re not wrong either. Cost is one thing to consider for-sure. Nobody wants to needlessly increase their burn-rate.

The problem for some is that some things just work better in containers (it’s easier to install multiple VM’s or containers and isolate over boundaries when transitioning a legacy application for example, perhaps limited infra budget also comes into play).

If certain developers don’t skill up now; there is also a feeling they may be left behind, or worse still at the mercy of vendors who are at present either established and have astronomical costs (impacting competitiveness) or are VC backed and at some point will die or need to start squeezing customers.

The free market already has problems with IT and knowledge at the moment, with many clinging to a single-vendor implementation or marketing material based knowledge. This is crippling decision making about going elsewhere, and in many cases about the products themselves. In 2014 my business had a regional vehicle and plant maintenance business approach us because their vendor (backed by MS) was over-priced, inaccurate and was holding any data they did have hostage…

Sure we have vague ideas of how to change things, but the opaque nature of the service-driven model leaves many high and dry, unless they get funding early, or they are lucky enough to have their problem defined by an open-source vendor. For many of the fringe I deal with, if they didn’t know me, they would be stuck paying vendor costs forever (It’s a completely different approach transitioning from open-source self-hosted or managed hosted to another managed hosting of the same software).

My feelings

  • Don’t inject complexity. Sure
  • Don’t expect to produce version 25 in iteration < 1,000. Sure
  • Don’t setup things nice-to-have before someone pays. Sure
  • Don’t use tools or understand the tools used to power your business…
    Doesn’t make a great deal of sense
  • Don’t increase your burn rate... 
    Well be sensible but know to pay for what you need to compete, and if thats a traditional SaaS platform, you can’t compete on features because they sell to everyone.

The trust-based model of vendors is dead. IBM, Oracle, Microsoft, Apple all screwed the pooch for everyone with contracts nobody understood; slow reactive products that basked in the ignorant glow of the market. That is why there is a huge backlash of DIY / in-house heroes.

Of course there are some people with budgets nobody can service. They deserve what they get, but for the most part, people don’t spend £4000/pcm if they could pay for a generic £2000/pcm that does what they need! People are balancing the cost of the unknowable and uncontrollable against the initially larger than they would like, but slower-growing, knowable, controllable, easier to innovate with. that’s solid business decision-making!

There are pain points for sure, and economies of scale are sometimes lost (although the open-source community at least partially compensates for this). Just consider that docker “for most” is one of the least complex things they need and they don’t need a whole enterprise architecture to use it.