Tom Dalling

I'm pleased to announce v1 of Tom's Data Onion — a programming puzzle in a text file. Let me know what you think! #TomsDataOnion tomdalling.com/toms-data-onio…

@carlchan1964 Thanks Carl! I barely use Twitter (X?) anymore, but this was a nice surprise to see when I logged in.

@tom_dalling Thank you for creating MirrorTheVideo.com! 💟 As a lefty learning to paint I can now follow along with ease and will paint like a pro sooner! My buddy who's wanting to learning programming; I will be sending him your way! Thank you so much Tom! 💟

@shimurokha @nateberkopec Yep

@tom_dalling @nateberkopec Sounds a lot like Hanami

We could have avoided 99% of this service object nonsense if we had just made Rails controllers easier to test

@nateberkopec Deprecating controllers specs in favour of request specs was a step in the right direction, although not everyone seems to have got the memo.

@nateberkopec Yeah, basically. An endpoint is an easily-testable function that takes a request and returns a response (streaming responses are a bit different, but still). Rails gonna Rails, though.

@allenholub Let’s have a look at Shi'er lü, the ancient Chinese music scale from before Big Pythagoras started spreading fake news. What a coincidence that it uses the same ratios.

@allenholub And what makes notes sound good together, Allen? Why are fifths always harmonic? Of course Pythagoras didn’t invent music. Do you also argue that gravity has no relation to math because Newton didn’t invent gravity?

Can anybody explain why so many (non-musician, non-mathematician) people believe there's some connection between music and math? I don't see one.

I don’t think he realises how much of a self burn this is.

@brandontroberts Probably “this should be reusable” (which btw is not feedback I’d typically give). IMO reviews are about the code, and less about who wrote it.

When you do code reviews, do you use "we" or "you" when proposing changes to a pull request? "We should make this reusable" vs "You should make this reusable" 🤔

@HashNotAdam Packwerk is exactly the kind of thing you want at larger team sizes, but it’s also going against the grain of Rails. DHH is a lone wolf and Rails is his personal “survival kit”, so it’s not surprising, just annoying that we have to rely on e.g. Shopify to make workarounds.

@HashNotAdam It’s less frameworks succeeding/falling, and more designs being better/worse. E.g. ActionController scales okish. There’s no reason why better design has to be less flexible. It should actually be the opposite. Bad design restricts your options, which is why AR is painful.

@MissRahee That’s the ideal, and if you can enforce that standard it will help. Even so, you can still run into the problem of important/risky things being deferred until late in the project.

@tom_dalling What you demo is meant to be an iteration that a real user could do something with. If it is then you'll naturally have to agree on and build the essential core first.

Continuing from last week: Against Must-Haves (Part Two) tomdalling.com/blog/software-…

@MissRahee Thanks Adam! In the article I was imagining that stuff was still getting done every cycle and being demoed. IMO demos show velocity, but don't necessarily derisk the project. Put another way: demos show that things are getting done, but don't ensure they are the right things.

@tom_dalling That was a great read and I'll be coming back to it. I don't believe agile is magic but which parts of this prioritisation problem are not solved by demoing a shipped iteration every sprint?

@nateberkopec Yep, but it at least gives control to the frontend, vs a typical REST API that responds with everything regardless of whether the frontend wants it. Some sophisticated REST APIs do give you that control, but they're not common and kinda complicated to implement well.

@tom_dalling The problem is that "extraneous data" is not always clear. If the client asks for 100 of X, filters out 99 of X, and displays 1, was that data extraneous? Well, is isn't if the filtering is implemented clientside.

I've grown a bit disillusioned by GraphQL. I hoped it would solve the problem of "kitchen sink" REST endpoints, but it seems to just have its own "kitchen sink" queries anyway. Maybe keeping state in two places was never going to work.

@t3dotgg So what do you see as your biggest factors that prevent bugs, if it’s not a test suite?

@bravelyjake Rock-solid CD is tablestakes for me. Our average time to production fix from report at Ping is under 7 minutes. Deploys taking more than 2 minutes is a disaster waiting to happen imo

Been almost 2 years since I wrote a test and I've never been more productive

@t3dotgg I don’t doubt you, but there is more than one potential explanation for this.

@GranataLLC @unclebobmartin @martinfowler “Recovery” or “rest” implies that you’re deliberately working at an unsustainable pace. If the pace was sustainable then rest wouldn’t be necessary.

NEW POST: Many software teams pack too much work into their iterations. Teams will usually run better when they have deliberate slack, as it allows their delivery to be more predictable and gives them time to improve their environment. martinfowler.com/bliki/Slack.ht…