Glen Mailer

Been doing this a lot recently, very satisfying when it comes to make the actual change: twitter.com/kentbeck/statu…

@jessitron Regression testing, but for the conceptual design of the system?

What if I had a "team member" responsible for worrying about two-faced cards? Every feature I implement, the agent asks this Feature Owner, "Look at this plan. Does it interact with two-faced cards?" It's this agent's only job to worry about that all day.

the invisible requirement to every new feature: "... and everything else still works" I bet we can do something new for this with agents.

@martinrue Best solution imo: stop using multiple domains

I had a CORS problem. I solved my CORS problem. So now I have a CORS problem.

@matteocollina @nodejs Are you able to share clear repro steps? I tried feeding that sample payload to an app of mine and I did not get a stack overflow or a crash - it was just rejected by Express - so I’m not sure what I’m missing or whether I’m affected

Today, @nodejs published a security release for Node.js that fixes a critical bug affecting virtually every production Node.js app. If you use React Server Components, Next.js, or ANY APM tool (Datadog, New Relic, OpenTelemetry), your app could be vulnerable to DoS attacks. 👇

Fiscal rules are relatively new. They didn't exist until the Maastricht Treaty in 1992. They spread across the globe following neoliberal politics. They are purely political in nature and are not based on any solid economics.

@simonw @obie I care a lot about how individual choices will lead to understandability and malleability over time I tend to work on long-lived software owned by small teams I’m not entirely anti-coding-agent, but I struggle to justify spending the time to really invest in learning the tools

This from @obie captures how I'm feeling about coding agents

@mitsuhiko We use ULIDs and this is a major reason I dislike them Imo, much like UUIDv7 they’re too big, don’t fit into URLs nicely, and aren’t recognizable enough Back when we just used autoinc you got to know the numbers for your most pathological customers - that’s much harder now

One UX quirk of uuid7 and tagged IDs is they the leading characters are low entropy so ID shortening doesn’t work. I think I dislike ellipsis in the middle. What are good options?

@jlongster @Vjeux I wonder if anyone now codes with their own preferred formatting and resets to prettier only when they’re finished

Great post by @Vjeux telling the story of Prettier's early days. Christopher was a big reason for Prettier becoming mainstream At the time I was starting a consulting business so after I wrote the first version that solved most cases, it was difficult for me to spend a lot more time on it. I did help with some of the early options like no semicolons (which I disagreed on) probably until May or so. I didn't have a salary job though so I needed to spend all my time making money through contracting and building my business. Grateful for other strong leaders/contributors that continued to build out and maintain Prettier x.com/Vjeux/status/1…

@simonw @mitsuhiko And it hasn’t even improved at being internally consistent within a page! Which is one of the core principles of react data architecture!

@mitsuhiko It's frustrating because GitHub was, for years, my go to example of why you don't need to go all-in on SPAs for a complex web application Now I guess it's my example for why you shouldn't give in to temptation and SPA something that works well already!

And while I'm ranting: I love react, but the GitHub react UI update is objectively a massive regression and I do not understand why they worked on that. What are they achieving by taking a perfectly working system and just making it worse?

@floydophone Sleact

cant believe i just built a streaming ui reconciliation framework targeting... slack messages

@ProCyclingStats the design tweaks look nice, especially on mobile - but I ran into a small bug: When filtering by team you don’t get the actual times anymore, you just get “ for everyone unless they were at the front of a group

@bostonou I’ve hit 100% of my deadlines in the last decade, it’s not that hard. The key is that almost none of my work has real deadlines, but when it does, we take them seriously

How to turn estimates into deadline: 1. Estimate happy path for all parts (~50% accuracy) 2. Find longest path of dependent work 3. Sum up that time and 2-3x it 4. Watch that longest path closely, which now has a large safety buffer

@martinrue getFullYear is not stupid, but I suppose there’s no good way to remove getYear that doesn’t break a non-zero amount of existing code

📖 Full article: martinrue.com/javascript-cho…

JavaScript Chose Violence (Link to full post in comments 👇)

@csswizardry Are you headed down for Frontend North?

@glenathan Probably not gonna get their round in, either 🤨 I’ll pop down soon and we can catch up though?

Developers: We know the bundle is huge but we fingerprint it and cache it for a year. Also developers: We release 17 times a day.

@csswizardry Yeah, that’d be good. I’ve got a 6 month old now so logistics are harder than they used to be!

@glenathan Rutland for one soon?

@csswizardry @ChaponFlo I am, yes!

@glenathan @ChaponFlo This conversation would definitely be easier with a pint. Are you still in Sheff?

@csswizardry @ChaponFlo That’s the idea, yeah - most releases should only touch a few files, and it tries to group them together through some heuristics node_modules should change rarely, for example

@glenathan @ChaponFlo Right, but the hash is deterministic: if the source hasn’t changed, nor will the hash. Are you saying webpack can rejig the source in a way that means chunks are less likely to change?

@csswizardry @ChaponFlo When configured correctly, webpack splits the bundle into a number of “chunks” each chunk contains a bunch of source files, and has its own name and hash. The idea is then that any change only invalidates a subset of chunks

@ChaponFlo @glenathan I must be misunderstanding. Do you mean ensuring that unchanged *source* files never make it into a revised bundle? Like keeping a record? Or the fact that unchanged *built* files’ hashes shouldn’t ever change anyway?

@csswizardry A correctly configured bundler is supposed to put the same files into the same named chunks, so most releases shouldn’t modify most chunks

@glenathan Compression dictionaries will fix it!

@csswizardry You can put the <file.txt at the start of the line too Although for most cases I still prefer the cat, I find it helps to think of everything as a series of pipes

Does it really matter? Probably not, but it’s a marginally cleaner habit, only spawns the one process, and if the file is missing, redirection won’t run pbcopy and you won’t accidentally wipe your existing clipboard.

If you want to copy the entire contents of a single file, don’t waste time piping cat into pbcopy—just redirect file.txt straight in!