Login
You're viewing the functional.cafe public feed.
  • Sep 18, 2026, 8:45 PM

    @jrconlin

    In my first few mechanical engineering classes, we were taught that real engineering is mostly pondering. That's what you are paid to do: understand the situation and think about it. That particular professor said your best work should be at your desk, with your feet up.

    It's why I think software "engineering" as we do it today is a joke: not enough pondering.

    💬 8🔄 56⭐ 156

Replies

  • Sep 18, 2026, 9:57 PM

    @jrdepriest

    One of the reasons I chafed at things like AGILE and other stuff was that, while the basic core ideas were solid, there were far, far too many ways that those ideas could be corrupted. "Points" and "Sizes" may be meaningless to us, but to anyone else, they're measurements.

    Likewise, call it what you want ("spike" or whatever) but by the end of the two week sprint, you had to have something to show for your effort that someone else can judge and critique you for. That does not lend itself well to thoughtful design work.

    🙎‍♂️"So, Bob, why did you spend half the sprint changing the design you specified last week?"
    👨‍🦰"Well, I learned a bunch of things last week that required redesigning those things."
    🙎‍♂️"That's not very efficient of you."
    👨‍🦰"Neither would spending several months designing something that doesn't work."
    🙎‍♂️"I don't see your point."

    💬 1🔄 1⭐ 0
  • Sep 19, 2026, 3:03 AM

    @jrconlin @jrdepriest

    I gather it's been demonstrated multiple times that open-ended, un-mandated research tends to be far more productive in the long run than specific, targeted research aimed at achieving specific goals.

    Unfortunately the latter tends to be favored by the people with the money to •fund• the research....

    💬 0🔄 0⭐ 10
  • Sep 19, 2026, 3:14 AM

    @cavyherd @jrdepriest

    Not really sure I agree, at least not within this context. Open research is different than directed problem solving. Most code is generated toward problem solving, or at least task resolution. That said, even those sorts of things need to be well understood before action is taken.

    If you're just adding a set of numbers, the problem space is pretty limited and sure, you can probably jump right to the code.

    If you're developing a new cryptographic library, things are quite a bit different, and you'd be far better off making sure that you really do understand the problem space and the sorts of ways that those things fail.

    Likewise, if you're building a markup language parser, knowing that there are probably already several dozen out there is probably going to save you a fair bit of effort.

    💬 1🔄 0⭐ 0
  • Sep 19, 2026, 3:16 AM

    @jrconlin @jrdepriest

    Fair. I was thinking research generally, not so much in a coding context. The latter does seem like a suboptimal context for "throw it at the wall & see what sticks."

    💬 0🔄 0⭐ 5
  • 💬 1🔄 1⭐ 0
  • 💬 0🔄 0⭐ 7
  • Sep 18, 2026, 11:50 PM

    @jrdepriest @jrconlin My uncle, who studied engineering at Durham, quotes his head of department as saying : "I am not here to teach you to be engineers. I am here to teach you to think."

    💬 0🔄 3⭐ 12
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 5
  • Sep 19, 2026, 5:16 AM

    @sigfpe I find that the past few months, I'm spending a lot more time pondering and a lot less writing code that I would consider routine and tedious. Because of Claude.

    💬 0🔄 0⭐ 2
  • Sep 19, 2026, 12:35 PM

    @jrdepriest @jrconlin thats kina why I believe my manager at work when he says that we are engineer. Well maybe we ponder too much...
    Sometimes we ponder all day

    💬 0🔄 0⭐ 0
  • Sep 19, 2026, 6:57 PM

    @jrdepriest @jrconlin
    I work as a so called "SRE", you can call it a glorified sysadmin.

    80% of the work when upgrading a major component (kubernets, postgres, clickhouse, what have you) is reading the changelog, and for each change, digging through our configuration repo to understand how we use the feature that changed and how the change affects us.

    Even then, it's still not enough pondering.

    💬 2🔄 1⭐ 7
  • Sep 19, 2026, 8:16 PM

    @wolf480pl @jrdepriest

    SRE folk are wizards who are able to bend mighty systems to their will. They are skilled jugglers of digital chainsaws and are woefully underappreciated.

    I frequently told folk that my job as a dev was to make sure that every SRE slept through the night, or at worst had at most two buttons to click at 3 AM if an alert ever went off.

    (I also noted that Test Engineers are the trained hackers that were nice enough to give us detailed damage reports. I loved getting stuff from them and giving them the opportunity to unleash holy hell upon my stuff. Any test engineer just running unit or integration tests was a waste of resources. Sadly, folk at the C level saw everyone as fungible because they had no clue what the skills of the job actually were.)

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 1
  • Sep 19, 2026, 8:53 PM

    @jrdepriest @jrconlin
    yeah but then you end up with 5 year old versions of stuff and never have time to upgrade...

    So I don't blame people who just upgrade on dev, wait a week to see if it blows up, if not, upgrade prod, without reading changelog at any point in the process.

    When the upstream project's rate of change is "web-scale", you're stuck between to bad options :/

    💬 1🔄 0⭐ 1
  • Sep 19, 2026, 10:39 PM

    @wolf480pl @jrdepriest

    I mean, a LOT of that supposes that the Change log is accurate. I think we've all see major, breaking changes rolled out in a minor or bug release.

    On a separate note, I noted that I would do version updates once a quarter (unless there was a critical CVE or something), and would do them by hand whenever possible. (There were a few engineers that said "Have Claude do that!", and it made a cold chill run down my spine.)

    Doing those updates re-familiarized me with the code. In addition, the changes were often subtle, and sometimes required changes in other parts of the library or code (so I didn't roll out some breaking change). Plus, they were kinda zen since it was an opportunity to learn some new approach. For instance updates to Rust generally included changes to  clippy which taught me new approaches.

    Surrendering that to some automated function basically made sure that the code was that less familiar to me, meaning that any future changes will probably have subtle errors or force me to learn the change anyway.

    💬 1🔄 0⭐ 0
  • Sep 19, 2026, 10:47 PM

    @jrconlin
    Oh, it doesn't, the second 80% of the work is fixing stuff after it broke due to things not described in the changelog, or that were described but you didn't thought would affect you.

    Also, version updates once a quarter? More like, one dependency a quarter.
    @jrdepriest

    💬 0🔄 0⭐ 2
  • 💬 0🔄 0⭐ 2