Filesystem Wars: Why Your Choice of Storage is Actually a Security Move
-1 points by vaguelytagged
-1 points by vaguelytagged
Strong LLM smell on this one.
So? The content is relevant, interesting and (to me at least) not an obvious rehash. It’s entirely possible the author did their own work and then had an LLM rewrite or polish it, maybe because the author isn’t a native Englishspeaker or is just poor at grammar or prose structure.
This flagging just comes down to “I don’t like the writing style”; if I were to use that metric I’d be flagging a lot of otherwise-valuable content here because I’m a writer and thus picky about grammar and such.
If you’re allergic to LLM prose, just skip the article. Why actively try to stop others from reading it too?
This is explicitly against the site guidelines now. Look at the definition of the spam tag.
Additionally, from a quick skim, there are a bunch of details it gets wrong, either by implication, omission, or just straight up hallucination.
In the About / Flags this is stated:
Spam which should be used for content that either is designed to promote a commercial service or for content that is created without meaningful human authorship.
As I understand it, it is not allowed to use the spam flag for cases when a human has written the content and then may have used the help of an LLM to improve the writing style (grammar, typos and such) or may have written the content in their native language and then used a translation service for translation.
…and actually I did not get the sense an LLM wrote this. It’s opinionated and makes some quirky jokes I wouldn’t expect from AI. It’s also lacking the easy tells like overuse of emdashes and lists. It reads to me like a human wrote it. And, as I said, the content is valuable; I read the whole thing and learned something from it.
Here’s what I don’t understand about the CoW approach to reliability: as described here, it assumes the changed sectors are written to physical media in the order the OS wrote them. If that’s true, then if a crash / power loss occurs at any point, the prior changes still leave the data structures intact.
But my understanding is that disk controllers don’t necessarily flush their caches in chronological order; more likely, they sort them by position in the media. That would break CoW assumptions. Or am I wrong on this, at least for SSD controllers?