kitl.ing
he/him foxdragon (bestboi), rust enthusiast.
29 posts
28 followers
28 following
Active Commenter
comment in response to
post
According to this video: youtu.be/L1NPFFRTzLo
the difference between gen5x16 and gen3x16 is 1-4% (on the 5090).
*That said*, what the framework appears to have is gen4x4, so half again as much bandwidth, and might start to make more of an impact.
comment in response to
post
It'd probably make the most difference in cases where the GPU is out of VRAM, and you're having to re-upload textures and such every frame.
comment in response to
post
It sounds like the full bandwidth isn't used most of the time, but makes a difference in latency for "the frame is waiting on this data" cases. In theory, a 4x reduction in bandwidth does not result in a 4x reduction in overall performance.
This is hearsay, and probably workload dependent too.
comment in response to
post
There's a PCIe slot (even if it's only a 4 lane slot) so you can run a dedicated graphics card if you want. Or possibly as an external GPU over USB4.
My question is whether the extra memory bandwidth helps close the gap on the extra cache provided by the X3D series of CPUs.
comment in response to
post
d'oh, i got the ending time of the auctions wrong. oh well. >_>
comment in response to
post
unfortunately, there's probably no legitimate way for me to qualify for the .th ccTLD, and .ith or .lith gTLDs ever existing seem kinda dubious.
comment in response to
post
And now, to an extent, I've come full circle, because I managed to snatch kitl.ing while i was setting up my bsky account.
comment in response to
post
A nickname that was used for me sometimes was "kitling", so when I started leaning into being furry a bit more, I started trying to use it as a username. But it's already taken in a lot of places.
So I started considering mutations of "kitling" and ended up with the more unique "kitlith".
comment in response to
post
A classic case of being able to bruteforce half as many bits, twice, is the 8 digit WPS pin. The way the protocol was setup, you could tell if the first half of the pin was correct even if the rest of the pin was wrong.
All because there was structure to exploit.
comment in response to
post
w3c-ccg.github.io/did-method-k...
comment in response to
post
if did:plc is "tied to a central db where you can rotate your key" and did:web is "tied irrevocably to this domain", did:key appears to be "tied irrevocably to this keypair." The draft spec calls out dids that will only used for brief periods, ephemeral actions and the like.
comment in response to
post
To be fair, this is ~2 levels of niche deep. You need to be interested in console hacking (awareness of atmosphere), and you need to be interested in what the components of atmosphere are (awareness of mesosphere). A lot of people will just setup atmosphere and not dig deeper.
comment in response to
post
Hah, FAQ dropped: one account per ID.
If you have alts, better choose which account you want verified wisely!
comment in response to
post
It's completely unrelated. Just happened to use angle brackets to indicate "placeholder"
comment in response to
post
In this context, it's a placeholder. I'm not naming the specific concept that I was thinking of when I wrote this.
comment in response to
post
Not really limited to russians, either. Just, "are there characters in UTF-8 that often look identical to ASCII characters" (yes), are they supported in punycode (yes), are there domain registars that allow you to register using those characters (yes).
Anyone can do it, under those criteria.
comment in response to
post
Not exactly. There would be example.com and xn--exmple-4nf.com (which, if rendered in a context supporting the decoding of punycode, would render as exаmple.com, with the cyrillic 'а')
The problem comes when someone is trying to visually distinguish the two. (i.e. a phishing link for gmail)
comment in response to
post
In theory, I think you might be able to traverse the component tree, and either modify the tree directly, or wrap the relevant function to add an additional menu item that way.
I'm running out of energy I'm willing to spend looking at this, though.
comment in response to
post
This looks like the relevant spot to add a menu item: github.com/bluesky-soci...
annoyingly, it probably shows up as "Anonymous" inside react devtools instead of having an actual name.
stackoverflow.com/questions/29... is potentially relevant for pulling shenanigans within a userscript.
comment in response to
post
From what I've heard, VRChat allows alt accounts, as long as you don't attempt to use the alt to ban-evade. So, using the same ID across multiple alts would not necessarily be defeating the purpose, as long as they are legitimate alts, and not two separate people.
comment in response to
post
What is NDMF doing here? Is there some kind of integration with UniGLTF/other SDKs that I'm missing?
comment in response to
post
As a green foxdragon, does that make me more 2000's or 2010's?
comment in response to
post
Found the message that partially states intent that I was thinking of:
bsky.app/profile/retr...
comment in response to
post
it's probably safer to transform (msg + generated pubkey) instead of just the original message, but I don't think it's strictly necessary.
That said, we're firmly in "rolling our own crypto" territory here. It's probably horribly broken in some way that I don't know how to spot.
comment in response to
post
The generated keypair here isn't the keypair used for "identity", so it should still follow the original requirements, aside from cheating by "recovering" the public key.
comment in response to
post
Depends on what the exact goal is. It only has to be derived in a way that's verifiable, so there's ways to shoehorn stuff into there. For example, you could generate a keypair, sign a transformation of the message, then derive the public key that corresponds to that signature and message+pubkey.
comment in response to
post
Is there a reason to use a random value instead of one derived from the message, aside from the same message generating the same key every time?