Comment by elric | Hacker News Reader

Comment by elric | Hacker News Reader

Parent Full thread

I've done quite a bit of government contracting for governments in Europe, including various government levels in Belgium and the EU itself. Every single time, I encountered a lot of friction regarding hosting. The applications were always hosted by a partner [1] that won a large public tender, usually by virtue of supposedly being the cheapest. Those hosting partners were invariably awful and incompetent. Simple tasks would take weeks. Complex tasks wouldn't happen at all. There was enough red tape to go to the moon and back.

Every project I worked on, developers (consultants like me) advocated for finding a way of bypassing the hosting partner, and either doing their thing on top of platforms like Hetzner, or switching to AWS/Azure. Not because they particularly love AWS, but because the default hosting solution sucks so much arse.

I can totally see EU cloud infrastructure working for the public sector, but ONLY if they come with the same level of self-service management as the big cloud providers. Without that, it's a non-starter.

[1] One of those is a large French company on the verge of banruptcy.

In my experience this is mix of lack of knowledge sometimes mixed with self-dealing/corruption.

But it can be prevented by writing hosting specs well. If in your proposed solution you specify really detailed hosting envinronment then most of these old incopetent companies wont even apply.

Unfortunately, the way my governments have dealt with the issue is by writing one very large tender, from which a Hosting Framework Agreement is derived. Individual projects have no say in it, and they're certainly not consulted when the agreements are drafted. So basically they're stuck with whatever is on offer, which isn't much.

A choice example was a hosted K8S solution: it was supposed to be highly available, but the framework agreement only specified 2 datacentre locations. The hosting provider thus couldn't build a reliable stretched cluster, and instead built two identical clusters. Forcing every project to figure out how to deploy their applications in a HA fashion on top of two clusters. Much fun. Not a giant waste of resources at all. /s

I have been working for large European utilities and even within AWS/Azure deploying something takes weeks.

Fairly certain this is true for large anythings if they haven’t been explicitly set up to avoid that problem.

Can you give a hint for [1] ?

Reply on news.ycombinator.com