Ancestors

Written by Glyph on 2024-10-28 at 23:03

Are you a software developer, and, if so, without looking it up, do you know what the THERAC-25 is?

=> More informations about this toot | More toots from glyph@mastodon.social

Written by Glyph on 2025-01-15 at 21:32

This poll is starting to get some attention again; not sure why, but it reminded me that I needed to attach this coda. https://x.com/zgryphon/status/1201292269787238400

=> More informations about this toot | More toots from glyph@mastodon.social

Written by Glyph on 2025-01-15 at 21:34

The summary here is that THERAC-25 is often described as a failure of testing or of concurrent algorithm design; i.e. a lesson for individual engineers about individualized engineering practices. And there are things to learn here, about those things. However, the deeper lesson is about liability structures, management practices, and cost cutting. i.e.: systemic explanations, not individualized ones. "Surprise, It's Just Capitalism™" strikes again.

=> More informations about this toot | More toots from glyph@mastodon.social

Written by Glyph on 2025-01-15 at 21:36

The code for THERAC-25 did have concurrency bugs, true, but why did it? Because it was written by one random programmer with no oversight and no accountability, because it was cheaper that way, because management thought they could get away with it, and nobody thought that software was important. Heaping blame on this (anonymous, apparently impossible to find) person misses the real blame that should have been placed squarely on the managers that allowed this process to happen.

=> More informations about this toot | More toots from glyph@mastodon.social

Toot

Written by Glyph on 2025-01-15 at 21:51

The call to action from this realization is that the original presentation of the THERAC-25 should get you to ask "hang on have we tested and documented this code adequately enough" but the refined understanding here should be to ask "is management giving us the resources we need to test and document this code thoroughly enough, are we making it a matter of policy that those resources always exist" and if the answer is no, form a union or sabotage the deploy pipeline or something

=> More informations about this toot | More toots from glyph@mastodon.social

Descendants

Written by Glyph on 2025-01-15 at 21:51

for no reason at all I'm just wondering if anybody at Meta is reading this today

=> More informations about this toot | More toots from glyph@mastodon.social

Written by Glyph on 2025-01-15 at 21:55

I have been lucky enough in my career to have pretty much always had management that was amenable to these arguments and actively participated in the development of good processes. hashtag not all managers, I guess. Some level of tension is even healthy here (engineers can absolutely become over-paranoid and just keep adding delays and process to avoid ever shipping anything; see https://blog.glyph.im/2023/12/safer-not-later.html).

But the trajectory of the industry makes me think that this is going to get harder.

=> More informations about this toot | More toots from glyph@mastodon.social

Written by tef on 2025-01-15 at 21:53

@glyph i mean, if they thought the sign applied to them, they wouldn't be at meta

=> More informations about this toot | More toots from tef@mastodon.social

Written by Glyph on 2025-01-15 at 21:56

@tef "This sign can't stop me, I can't use mastodon!"

=> More informations about this toot | More toots from glyph@mastodon.social