> systems aren't maintainable except by people that are very familiar with them. I think that a consequence of "two sorts of knowledge: techne and metis" is th

> systems aren't maintainable except by people that are very familiar with them. I think that a consequence of "two sorts of knowledge: techne and metis" is that standardisation is good, but it only gets you so far. Past that point, you need to be familiar with the system. This should not devalue our efforts to standardise, e.g. get systems to all log to the same aggregator, and emit the same basic stats, agree on naming and forwarding of correlation ids that will allow us to cross-reference related log entries. But we should also recognise that those efforts will never cover everything. e.g. If I changed over to working on an unfamiliar system in the same organisation, I would know where it should be logging to, what the field naming and general structure of those log entries should be, but I would not not know what healthy operation should look like in those logs.