Going to give publishing weeknotes a try. Weeknotes are something that have been quietly popping up, and they’re essentially a journal-style blog post reflecting on a given week. They can be as personal, or impersonal as the writer likes. There’s no fixed cadence, so missing a couple of weeks here or there is no biggie, but they’re great ways of reflecting on a week. I’ll be starting by taking some of the ad-hoc notes and links I normally jot in Bear throughout the week, and running from there.
Speedcurve and WPT are fantastic for debugging problems, but Puppeteer on its own might be an even better Swiss-army-knife for improving WebPerf (especially in local environments): Michał Janaszek has a great article on the subject with useful examples and I love this candlestick chart comparing different scenarios against different network states.
Maybe obvious but still important: Saying problems are “hard” or “difficult” to a person sounds like we’re being lazy. Better to explain the reasons why something is hard, framed in a way that makes sense to that particular person.
Riiiiiight, it’s ITIL time! An Incident is disruption to service [whether customer impacting, or not], a Problem is the cause of multiple incidents. A good analogy for complex systems is that of symptoms and causes: Having a runny nose is potentially a symptom caused by suffering from the flu, and the symptom can be treated with certain medicine—A runny nose is an incident that can be solved, but the underlying problem of a flu will persist.
“Python is the second-best language for everything”: This is an old saying which has stuck in my head for a while, recently resurfaced during Dan Callahan's Keynote at PyCon2018. As far as languages go, of the few I’ve dealt with, it’s not one I have particularly enjoyed, and so I’ve shied away from really learning and using it properly. And yet quote still really rings true. Between all these complicated provisioning shell scripts with snippets of awk, I’d much rather see them written Python because it’s imperative, syntactically consistent and easily testable: More approachable to an outsider. I think I should invest more time in learning Python.
David Maidment's talk at Thursday's BrumPHP on "Handling Exponential Growth with ElasticSearch" was an interesting and refreshingly honest insight on how a tech stack may already hide the answer to new problems, as well as some hard truths about how "better" performance can already be enough. You can watch it here.