i rly love this 2022 grothendieck profile from the nyer https://www.newyorker.com/magazine/2022/05/16/the-mysterious-disappearance-of-a-revolutionary-mathematician now that i understand grothendieck's work more, and now that i feel very strongly that i understand some of his state of mind as an OCD lik mine, it has more power to me
44
Replies
·
0
Boosts
i also think his work on differential geometry (and of course not just him, but the whole field) is relevant and important to apply to filesystem design. so i am hoping to find ways to learn about sheaves and rings and groups and other objects that i can manipulate to achieve the semantics i want
@hipsterelectron ooOOoh!
...
Well, FSs just got more complex for me! /s
importantly, filesystems (and in fact all data and memory) have more than one set of desired semantics, depending upon use case. the most basic one is understood as "volatile" vs non-volatile memory, where non-volatile ("persistent") data is written such that it can be retrieved elsewhere, outside of the working memory of a computation (in particular, even after power is removed from the system). but there are a multitude of access, availability, error-detection, error-correction, and other goals and expectations we (users, like scientists) will have for that data. and these semantics are all mixed together by all extant operating systems
this is, in my view, most directly demonstrated through scheduling and timing—in particular, control over how tasks are scheduled. a "timing attack" which reveals the contents of secret data is possible because the scheduler is in fact controllable by an attacker—even remotely. i claim this is a fundamental problem that is completely fucking ignored by the entire field of cryptography
i am aware there is much research on timing attacks in cryptography. without fail, this research advocates for the use of algorithms that claim to mask data-dependent behavior in the cpu itself. this is, to put it bluntly, never effectively demonstrated in a scientific manner, at any level of cryptographic research.
