@rauchg@vercel@nextjs Love seeing how much value you get from feature flags! One thing Iβm curious aboutβhow do you manage the technical debt side? Flags tend to accumulate over time. At @getunleash, we see customers creating 2 flags for every 1 they remove. How do you keep things clean?
Account sharing might be fine for less critical systems, but when it comes to a feature flag platform, individual accounts are essential.
https://t.co/6bxoi9R9wr
Picked up a great habit from my colleagues at @getunleash: using Grafana/VictoriaMetrics to guide day-to-day development. Spent 20% of this week figuring out what to visualize, creating metrics, and setting up charts to showcase the impact of our deliverables.
@ivarconr Thatβs how, as an industry, we got misled by story points, burndown charts, and other effort or money-spent metricsβspeeding up from 100 to 200 km/h, only to realize weβre going really fast... but in the wrong direction :)
One of the most satisfying results of impact mapping? Addressing behavior changes with deliverables that donβt require writing a single line of code. Winning at software development is all about outcomes, not output.
Honored to see our work on @getunleash acknowledged by the @thoughtworks Technology Radar! What makes this especially meaningful is that you canβt buy your way into the Radarβitβs based on real practitioner insights. π https://t.co/7GWKTRcj7I
The recent #CrowdStrike incident reminds us why companies should seriously consider progressive delivery at deploy time (canary deployments) and at runtime (feature flags).
My code goes to production almost every day. I love the confidence from my progressive delivery setup:
* Progressive deployment (canary deployment with @kubernetesio release channels operator)
* Progressive release (feature flags controlled releases with @getunleash)
When updating components behind a feature flag, it's best to rename the old component and retain the original name for the new development to preserve git history.
Forget about traditional WebPerf metrics like LCP, TTFB, TBT. After skimming through https://t.co/ZZ5AqQwIxJ, I've come up with a new metric for a bloated JS bundle:
BTP - Bigger Than Pornhub.
My preferences in terms of inherited code:
1) Code that was anticipating likely vector of change
2) Naive code that was not anticipating any change
3) "Extensible" code that was anticipating every possible vector of change
Oh my - weβre getting a little famous! π
Unleash, the leading Open Source Feature Flag platform (thatβs us! π) just hit 9k stars on GitHub at
https://t.co/4tj9PAg937
Thank you #Unleash community!