vinay_ys

I didn't know chrome didn't allow self-signed certificates. Since when?

Our society has already accepted many such evils and you are the beneficiary. We breed/rear sentient animals in very poor living conditions with explicit intent to kill them and eat them. We conduct experiments (clinical trials) on humans. These experiments have significant risk of harm. We also sell for profit, medical treatments and commercial products with significant risks of harm. We make them life-long-dependent or addictive so that our profit flow doesn't stop. We sell weapons/machinery of mass destructions for profit and we foment conflict to ensure we have proving grounds and strong market for those weapons. We externalize costs from our own society to faraway poorer societies for our own selfish interests. Point is, a few elites make moral decisions for all the masses and many times in history we see how these elite decision makers had no skin in the game. So, stop being judgmental. And make scientific arguments backed by data and context if you want to truly improve things.

Judgmental comments like this don't serve open and honest intellectual discussion. It is full of moral judgement and shaming meant to suppress the other side. It only shows this commenters privilege to hold luxury opinions at the cost to others. We should learn from the moral policing that happened w.r.t covid discussions that lead to poor decisions with bad science.

The argument that combining traditional database and vector database into one because it reduces data movement doesn't compute for me.

Firstly, even for non vector data, read/write transactional database vs read-optimized store purely for fast serving are already markedly different. Then, the shape of data that is used to generate embeddings is markedly different than the shape of data that is ready to transact or serve.

So, no matter where it is stored, it has to leave that store, get transformed and enriched and then run through an embeddings generator (ML inference).

Then, it has to be stored in a manner that is optimal for retrieval ranking. If you are doing ANN that's one thing, but if you are doing attribute based filtering while retrieving and you wish to accelerate it through GPU to do more exhaustive search, that's another thing altogether.

All these lead to fairly sophisticated optimized implementation. Sure, a singular database product solution that has all these different optimized engines can emerge over time but surely it is too early today to converge like this.

Traditional corporate network architectures were electrical packet switching networks with a hierarchical topology. This design worked well for smaller networks, but when scaled to large datacenter networks, it resulted in unpredictable packet blocking and dropping. This made it difficult to build disaggregated storage-compute architectures for latency-critical distributed SQL/NoSQL databases or near-realtime streaming big-data workloads.

The advent of faster and cheaper network speeds and feeds made it economically feasible to recreate decades-old circuit switching network topologies like Clos/Fat-tree in packet switched networks. This solved the problem of packet blocking and dropping for any-to-any traffic patterns, even at line speed. This simplification of the network topology made it easier to place workloads within the datacenter and to build larger workloads.

However, there was still a problem of head-of-line blocking due to flow routing imbalances. This could cause packet drops and latency spikes. This problem was solved with the development of end-to-end traffic aware routing techniques using software defined networks (SDNs).

As companies built more datacenters, they needed to interconnect them with high-speed campus/metro networks. These networks could be fully meshed, but there was a limit to how much bandwidth they could support. This led to the need for software defined controls for managing traffic priority and quality of service (QoS) across these high-demand links. This was especially important when campus/metro links failed and degraded capacity.

However, implementing all of these network functions in routers would have made them very complex, expensive, and difficult to manage. This led to a deeper dive into SDN.

As companies gained more experience with SDN, they realized that the traffic patterns between large pods of hardware (cells) within a single datacenter were fairly stable. This meant that the full-clos topology between these cells was underutilized. Additionally, the next generation of network speeds and feeds was so fast that it would have required a complete rip and replace of the existing infrastructure.

To avoid this rip and replace, companies looked for ways to make the existing layers feed/speed agnostic and to improve their utilization. This led to an interest in optical switching, but with much faster switching speeds and software defined control (unlike traditional telecom optical circuit switching).

Today, the hardware and software stack needed to implement all of this is still very bespoke and not as mature as electrical packet switching networks. This means that it is not yet a viable solution for smaller players. But the amount of investment AI/ML workloads are attracting and how much of a big difference these OCS interconnects can make between these large pods will likely lead to renewed interest in commercializing the OCS solutions for smaller players. So, in that way, this publication is very timely.

Well, not exactly. They are making privacy improvements in open standard protocols that you are free to use without using their service.

ECH is still not working with Cloudflare in all regions. If someone from Cloudflare see this, please let us know the status of ECH rollout in all regions.

Chrome isn't a not-for-profit charity project. We know it is a free (not paid for by the user) software built by a for-profit company and we can safely assume the intention to monetize it exists. Even so, I'm glad Chrome exists as it has meaningfully accelerated web technologies forward and has helped keep web secure. In return for their investments, if they get to make money off it in clear and transparent ways, I don't have a problem with it.

We act as though we are finding out the ads powered free software services business model today. All of user facing google products – Chrome, Android, Gmail, Youtube, Photos, GSuite etc all exists because they are paid for by ads revenue from ads side of the business. One can make a very strong case that the net benefit far far out weights any harm on the other side.

Ah the custom keyboard leak vector. I remember this controversy now.

I wonder if using less of memory-recall capabilities of the brain will lead to loss of other important capabilities. So, I worry if using too much tech crutches is going to be detrimental.

I seriously hope this doesn't take off and fails spectacularly like Google Glass. I don't want a society where the people I'm interacting with are always recording everything around and giving it to a private profit corporation who can change their minds about what they will do all that private data any time they like.

Interesting. Totally missed this news. Could you edit/post some links to this coverage?

Today, every telecom company is also a OTT apps/services company with desires to become like big tech. So, your assertions about telecom companies as some sort of time frozen thing is not valid.

Secondly, I would trust Google more than Apple or Meta to keep my data safe from technical hacks or leaking or selling it to third parties. And I understand they are going to use my data to power their ads business where they personalize and target ads to me on their own and 3rdparty properties.

Finally, we need to keep the regulatory oversight and pressures to move and keep them in line with increasing privacy and security expectations.

Either use a full-fledged library like tcmalloc or jemalloc with profiler and debugger capabilities built-in or do tooled generated code that's ensured to be bug-free and usually tied to lifecycle of request/response or connection etc (or game level, scene, session etc). Hand-rolling like this in 2023 is neither necessary nor optimal.

I wish Siri did that consistently. Right now my Apple Watch Siri is unusable even to set timers. I don't know if the latest updates on my old watch struggles to run or what else the issue could be.

This seems like a hack. why doesn't mail channels do a proper authentication of the sender?

To be clear, this issue is really with mail channels not authenticating the sender as the verified owner of the sending domain. CF workers are aren't the issue here.

Yep, I see phishing style spam (email from random domain that claims to be my bank) in my gmail in the past few weeks.

So, they think they can stop ransomware attacks by hiring someone willing to work 10 hours shifts 7 days a week? Good luck with that!

You can have lots of threads – they are just thread stacks holding execution context – but that doesn't mean they will run well. You also need lots of execution units on which they can run. Even then, if any thread can sleep/wake-up at any random time, you will have a lot of jitter in your thread response latency. So, such large number of threads is anti efficient or predictable performance. It is not about that at all but instead it is about programming style/habits.

What these 0days teach us is that if you are a person in the crosshairs of powerful adversaries you have to be super-paranoid and reduce your target surface to as small as possible.

If you are James Bond and you want to have secure communications with M, you are better off using a custom appliance type mobile device meant only to do that one function and nothing else. You will have to rely on anonymous 3rd-party burner devices to leave messages encoded in pre-agreed sequence of rotating geo-locked codebooks on random message boards. Any less opsec and you are toast!

If you are a normal human-being using connected digital devices normally, then you should assume anything you put on those devices is already stolen. If you want to keep things truly private, keep it off digital (paper, old-school analog tape etc). Then, at least they will have to physically steal it from you – which, depending on the situation, might be much harder for them, but not necessarily much safer for you. So, you lose either way.

Only real way out is to have well-functioning democratic governments with strong transparency and checks and balances with strongly enforced privacy laws to protect its constituents from both domestic and foreign adversaries and strong support for initiatives like Citizen Labs.

Browsers only enforce that certs were logged to CT logs (because they will fail a TLS connection unless a certificate has valid SCTs attached to them). The actual domain owner will have to monitor the CT logs and call out when they notice a certificate being issued that they didn't request. Without that active monitoring of CT logs by the domain owner, it won't help.

In India, even before we get to cousins, we have distinct terms for father/mother's siblings of same and opposite genders. Same gender elder/younger siblings of your parents have different pronouns that translate this way in Kannada (Language of Karnataka state):

senior-father - ದೊಡ್ಡಪ್ಪ (doddaappa) for elder brother of father,

senior-mother - ದೊಡ್ಡಮ್ಮ (doddammA) for elder sister of mother,

junior-father – ಚಿಕ್ಕಪ್ಪ (chikkappa) for younger brother of father,

junior-mother - ಚಿಕ್ಕಮ್ಮ (chikkammA) for younger sister of mother.

Their spouses get the same seniority title with appropriate gender.

Then, opposite gender siblings of your parents get pronouns that are more conceptually closer to uncle/aunt. But here, there's no seniority distinction.

Even with an honest company, the pressures on them are twofold – security and legal. Their systems can be compromised through security vulnerabilities and social engineering (including coercion – money, ideology, compromise, ego – classic psyops playbook). Or they can get legal government orders - which pretty much every government in the world have laws on books and operational practices to force any actor to hand-over data in the name of fighting money laundering and terrorism (AML/CFT). It is very expensive to put up strong defenses against these. I don't see a viable business model charging $5/month that covers regular operational expenses and covers these types of events.

Edit: Forgot to mention backdoors built into basic technologies they may already be using – like the Cavium HSM thing that came to light earlier this week.

Most of USA doesn't speak English as their first language, isn't it?

Why are we still talking about broken notions of immutable systems when we actually have trusted execution environments and secure boot and cryptographically sealed strong assurances of what is executing at any point and can do secure upgrades, even with remotely attached secure storage and do it across a huge fleet of machines at very large scales?

You sound like James Burnam.

I'm waiting for a well-optimized LLM-based system built-into a local editor like obsidian and I can ask it scan my entire local Documents folder and then it supercharge my reading/writing locally.

> A couple of troublesome example IPs are 193.142.125.129, 129.250.6.113, and 129.250.3.250. They come up in a UK traceroute - and I believe they're in London - but geolocate all over the world.

If I'm running a popular app/web service, I would have my own AS number and I will have purchased a few blocks of IP addresses under this AS and then I would advertize these addresses from multiple owned/rented datacenters around the world.

These BGP advertisements would be to my different upstream Internet service providers (ISPs) in different locations.

For a given advertisement from a particular location, if you see a regional ISP as upstream, you can make an educated guess that this particular datacenter is in that region. If these are Tier 1 ISPs who provide direct connectivity around the world, then even that guess is not possible.

You can see the BGP relationships in a looking glass tool like bgp.tools – https://bgp.tools/prefix/193.142.125.0/24#connectivity

If you have ability to do traceroute from multiple probes sprinkled across the globe with known locations, then you could triangulate by looking at the fixed IPs of the intermediate router interfaces.

Even this is is defeated if I were to use a CDN like Cloudflare to advertise my IP blocks to their 200+ PoPs and ride their private networks across the globe to my datacenters.

Let's not throw the baby out with the bathwater. Distributed horizontal scaling is a necessity at any non-trivial scale. Yes, mindless cargo cult excessiveness of anything is bad. The real crime here is forgetting how to organize large codebases effectively. We used to have beautiful software frameworks that allowed things to be developed in neat little modules with interface bindings that can be turned into in-proc method invocation, or IPC invocations or RPC invocations through compile time switches to create one or more deployable binaries with neat little infrastructure for service discovery and routing, end to end tracing and profiling etc. The modules were independently verifiable for correctness and performance. Modules had semantic contracts clearly defined and their implementations could be changed transparently. We had sophisticated distributed persistence layers with MVCC, zero-cost snapshots etc. with remote addressable storage 20 years ago. Instead of scaling these patterns for web-scale with the necessary engineering discipline, we took a lot of cheap shortcuts only to pay for them 10 times over in the form of wasted dev time and hardware.