kevvok

This issue is already being tracked by a shiny website: https://tldr.fail/

It’s not implemented in the Linux kernel, but the latency penalty you’re describing is part of the “Happy Eyeballs” algorithm: https://en.wikipedia.org/wiki/Happy_Eyeballs

It’s pretty common for IETF drafts to be substantially complete well before they are finalized as RFCs. For example, supporting ML-KEM in TLS is still a draft, but there are already multiple large scale deployments of it since the technical aspects were nailed down a while ago

Also missing Band 13, which is Verizon’s main band for coverage

I think that'll only happen when and if the corresponding drop in sales offsets increases in revenue from the subscription services owners will be forced to use. When they announced this originally for EVs it was clear the underlying motivation was to convert owners from a one-time source of income into an ongoing stream by forcing them into a subscription model for features they would get from CarPlay/Android Auto.

The smaller switches like the Arista 710P are meant for deployment out at the edge of the network where you want something small and quiet (e.g. at people’s desks or in conference rooms) to provide more ports without needing as many runs back to the network core where the big loud switches live. They’re still enterpise grade since they support enterprise features like centralized management, VLANs, QoS, IGMP snooping, etc.

Yeah, the ClientHello message(s) are already part of the transcript signed by the server (and the client in mTLS) during the handshake

This reminds me of how Sarah at the Connections Museum in Seattle wrote a driver to allow an Asterisk soft PBX on a Linux box to speak revertive pulse signaling to the pre-DTMF trunks on their old telephone switches

https://www.youtube.com/watch?v=35N5vKKGDy8

Years of middleboxes tampering with TCP and barfing on anything they didn’t recognize is why QUIC encrypts everything as well

Many operators do configure the SIP signaling for VoLTE to use an IPsec transport terminated at the P-CSCF, but most (if not all) of them only configure IPsec to provide integrity protection.

Murthy Renduchintala

He was a joke at Qualcomm before he went to Intel too. That Intel considered snagging him a coup was a consistent source of amusement.

With the industry pivoting towards focussing on post-quantum algorithms, I’d be surprised if yet another elliptic curve gains much traction.

I’ve heard this about TOMS

Oh don’t worry, they have plenty of those too

But I'm not sure if it works the same way against elliptic curves - or at least you'd probably need a bigger computer to attack the same level of security.

Elliptic curve cryptography is also based on the difficulty of computing discrete logarithms, which makes it vulnerable to Shor’s algorithm. Unfortunately, while the increased difficulty of brute forcing ECC with a classical computer allowed it to use smaller key sizes to achieve security equivalent to older algorithms like RSA, the smaller key sizes make ECC attackable with fewer qubits.

The point was a prefix from 2600::/16 isn’t special; it’s just from one of the blocks assigned to ARIN. One ISP I know of with an allocation in that range is Verizon, which announces a number of prefixes over BGP in 2600:1000::/24

According to the IANA assignment list, 2600::/12 is assognrd to ARIN.

https://www.iana.org/assignments/ipv6-unicast-address-assign...

Hell, even the JWST is powered by the same CPU as the first iMac running at half the speed

There’s a great book about the development of Cell and Xenon (which went into the Xbox 360) called “The Race for a New Game Machine” that was co-written by a couple folks who worked on them at IBM.

Lexus also has so-called "noise-reduction wheels" that utilize resonance, though somewhat differently: https://elevatingsound.com/how-the-noise-reduction-wheels-of...

The use of IPv6 in cellular devices is specified by RFC 7066. When a device first registers for a data connection, the network provides an interface identifier to use for the lifetime of the registration, which is guaranteed not to conflict with the one used by the router. The network also allocates a /64 prefix exclusively for use by the device. These two things allows the device to perform SLAAC for link-local and routable addresses without needing to do duplicate address detection.

It’s also worth mentioning SLAAC does not strictly require a hardware address to function. For example, Apple devices implement newer standards from IETF to generate random but stable addresses that don’t reveal information about the hardware addresses (https://support.apple.com/guide/security/ipv6-security-seccb...)

Radio firmware and state machines have always had weird bugs, and even when it conforms to standards (some of which are extremely interpretable), does very weird things in the real world. Pre-smartphone, being able to update phone and radio firmware was extremely rare, so it was common for the networks instead to implement workarounds on a manufacturer or handset basis. Having a hardware ID that identified this was extremely useful

Now that it’s common for devices to be updated regularly, they will typically send an extended form of the IMEI to the network called the IMEISV, which is the same as the IMEI except the final check digit is replaced with a two-digit code indicating the current software version (SV = Software Version).

You’re right that the signals have to be converted back into bits at the destination. Basically, this solves the problem of pumping those bits at high speed across traces on a circuit board vs within a chip. The longer a signal has to go, the harder it is to maintain its integrity.

And presumably ‘Gesserit’ is meant to be a distortion of ‘Jesuit’

They’re just on T-Mobile since US Cellular dropped them last year

Same here: this video helped me finally understand the basics concepts underlying LWE

Interestingly, a CVE has been assigned for this issue: https://www.cve.org/CVERecord?id=CVE-2023-49298

Has Qualcomm ever engaged in significant semi-custom work?

Yes, the early generations of Snapdragon SoCs were built using a custom ARMv7 core called Scorpion, which was later succeeded by another v7 core called Krait. The first ARMv8 SoC (the infamously disastrous Snapdragon 810) used stock ARM cores, but the next generation (820) switched to an in-house v8 design called Kryo. Around that time though, the company did massive layoffs, including in silicon engineering, leading to all the subsequent designs using tweaked ARM cores (despite also being called Kryo)

Algorithms using elliptic curves can also be broken using Shor’s algorithm

Even just XORing the address with a hardcoded, pseudorandom bit pattern would have alleviated the issue.

Yeah, that's in the same vein as Seznec's 'skew-associative' cache where a cache line would be either direct mapped or mapped based on some hash of its address