4 ms·
Thank you, very interesting insights that I hadn't thought about. I think you're right that the essential difference is complexity, and I am taking a "from the
by snowwrestler 11y ago
Thank you, very interesting insights that I hadn't thought about. I think you're right that the essential difference is complexity, and I am taking a "from the top" view.
It makes me think of the various disciplines within science. In theory they are a single stack--chemistry can't violate quantum mechanics, because chemical interactions are just very complex quantum mechanical interactions. But they're so complex that the mathematical tools of quantum mechanics can't be practically used for chemistry (beyond super simple reactions). For humans to analyze and manage chemical reactions, we had to invent a new set of analytical tools that are specific to the level of abstraction that we call chemistry.
And again, on top of that goes biology, which is after all just very complex chemical reactions. But we don't use chemistry equations (or quantum mechanical equations) to analyze genetic disorders.
Maybe software is sort of akin to biology. It's just super complex circuit design, which is itself complex applications of basic electromagnetic theory. On the way "up from the bottom," each time the complexity gets to be too much, we have to invent new mental tools to manage the system at a new level of abstraction.
Are the mental tools for software as robust as they are for circuit design? I think about the implications of security when it comes to APIs and packages. I might not want to update my package, but what if a major vulnerability is disclosed and patched? Now I have to update it, and because security and features are often pushed to the same production branch, I might get breaking changes with the patch. I know not every API and package does this, but it seems like many do. Old release versions are only supported for so long, typically. And depending on the patch, it might not even be possible to maintain complete compatibility with old APIs.
I agree with you that the key is culture. I've felt a significant shift in managing websites over the past year due to security. Instead of days to apply open source community patches, I only have hours, or maybe minutes. The default is shifting from "test security patches before deployment" to "deploy security patches instantly then see what we need to fix," because it's harder to recover from a hack than to adapt to documented breaking changes.
I think the implications of security are only just now beginning to percolate into popular consciousness, which will create demand for higher standards with software "engineering." Look at the recent car hacking stories, or the Ars Technica coverage of IoT security, which got picked up by NPR among others.
Thanks for good discussion.