jsolson

Speaking for the depraved, I think the obvious questions are how little I _must_ drink, how much I _must_ exercise, etc.

Ideally it would be "uncorrelated" and "none"

Seems unlikely, though

Within the US, with a modest amount of pre-pandemic international travel, WhatsApp was a thing I was aware of, but never had cause to use (or even install). I'm not certain I've even read the name since the pandemic started before today.

Everyone I know uses some other platform, or just SMS because it works.

As others noted, it only supported horizontal walls and floors. More specifically, though, it only supported _one_ gap (e.g., a single window or door). You couldn't stack floors on top of each other.

This allowed a single ray cast against the 2D map to learn which vertical slices of wall texture to paint (no pitch or roll, remember). Floors and ceilings could be painted in vertical order (above / below the horizon line - "floor" and "ceiling" being explicit for any given open area), iirc, and walls filled in based on the ray cast from the eye.

Below a certain level I mostly agree with you. One company in my past didn't work this way, and one too many pages in the middle of the night for a business I wasn't personally invested in made it an easy job to walk away from.

That said, there is a level of seniority at which "their problem" is "your problem." If you'd chuck the business out the door rather than getting paged a few times a year, you're not ready for that level yet. This varies by company and organization size, but fundamentally you can't (and shouldn't try to) anticipate everything up front. Sometimes you need the knowledge, judgement, or simply signing authority (literal or metaphorical) of someone specific.

Protobuf supports sum types in the higher-level generated descriptors and languages -- on the wire they're just encoded as, well... oneof a number of possible options.

Firecracker relies on KVM in the Linux kernel which relies on VT-x or similar. Offering VT-x inside a VM that's already using VT-x is called nested virtualization. EC2 VMs do not support nested virtualization, while others do.

That said, it works fine on EC2 Bare Metal instances.

(n.b., I work on the virtualization stack on Google Compute Engine)

Awwww, I kinda like Buganizer -- but also it allows for much more direct engagement between Googlers and the issues reported by the public.

For my personal work, I treat GPLv3 (and similarly licensed) projects as essentially the same as anything behind a commercial license -- they don't exist.

Rather than thinking about how to re-implement a thing that doesn't exist, I think about how to implement the tool I actually need. What I end up with may be less general, but it's actual functionality is often simpler and easier for me to grok/remember.

Without going into details that would spoil things, the TV series "The Good Place" touches on this idea, although it's Madden rather than Mario.

Well, if it helps, fixating on this choice allowed me to collapse your entire thread without reading most of your comment with zero guilt :)

> I prefer the known, extremely low risk of contacting COVID

I'm a little confused by this. I live in King County, WA (and I work at Google). According to data reported by the county (as tabulated by the NY Times), one in 19 residents have contracted COVID[0]. That's >5% -- with a 1.8% case fatality rate in the US[1]. That's a bit below a one in 1000 chance of having died of COVID if you live in my county, including the infection fraction. If you look a the risk of just contracting it, based on the county, the risk is higher than any other disease I'm aware of.

Am I missing something?

[0]: https://www.nytimes.com/interactive/2021/us/king-washington-...

[1]: https://coronavirus.jhu.edu/data/mortality

If you're curious, The Secret Life of Groceries is a good (and interesting) read.

One tl; dr thing is that the supermarkets are a _supplier_ to both end customers and those with products to sell. They supply shelf space and customer reach, in the very literal sense that companies bid for things like endcap placement (the end of the aisle being better than being in the aisle, since everyone making an orbit through, say, the deli section will pass your goods).

+1 to congrats -- my read on this:

- There's a bunch of RJ45 up top that I don't quite understand :)

- A bunch of storage sleds

- A compute sled, 100G QSFP switch, compute sled sandwich

- Power distribution (rectifiers, I'd think, unless it's AC to the trays?)

- Another CSC sandwich

- More storage.

I assume in reality we'd have many more cables making things less pretty, given the number of front-facing QSFPs on those ToRs.

Going a bit off-topic here, but as a tiramisu lover I feel compelled to say that if you like desserts and coffee, generally, you might consider trying some if the opportunity presents itself.

A bit more on topic, I'd generally agree that it's not so common as to be universal. I don't think I'd heard the word in any way that registered as significant before my mid-20s, and I'm certain I never tried it until my late 20s or early 30s. I grew up in the suburbs around Pittsburgh, PA with plenty of passable Italian restaurants -- but tiramisu simply wasn't a dessert people in my family ordered.

Taking a glance at levels.fyi, in NYC for people with one year of experience there isn't a single entry from Google below $150k. Similar for Amazon.

It's not an influenza virus.

That said, even if it was, a "very strong flu" with global transmission and impact _is_ a pandemic -- see the 1918 Flu Pandemic.

The death rate in 2020 in the US also surpassed that during the 1918 pandemic.

What I think we've learned is that the same measures used to slow the spread of COVID-19 are staggeringly effective against the flu, and that we now know how to prepare for the next particularly aggressive influenza variant.

At Google it mostly _is_ your responsibility to do that, yes.

There is substantial tooling assistance to assist with this, and it's common to make changes by adding new functionality, writing an automated transform to shift usage from old to new, sharding out the reviews to the suitable OWNERS, and finally removing the old functionality.

Very heavily used common code tends to be owned by teams that are used to this burden. That said, it does complicate upgrades to third party components.

There have been some changes that make this entirely untrue for earlier promos and partially untrue for promotions to L6/Staff. There's considerably more locality at this point.

> Just as you can’t go back to buying a prebuilt PC if you’ve built one in the past.

Eh, I think it depends on where your life takes you. Other than an IBM PC XT, I'm not sure we ever had a pre-built desktop PC in our home while I was growing up. Naturally I built all of my computers going off to school, but around ~2004 I made the jump to mostly Macs (and later Chrome devices) -- since then the closest I've come to building my own PC at home is buying an Intel NUC kit and outfitting it with an SSD and some RAM. That said, I think I get my fill of computer tinkering at work fiddling with BIOS and other firmware SPIs, PCIe cabling, bad QSFP optics, and even the odd trivial PCB rework :)

On the other hand, while I don't see it as likely I'll get into biking, I own enough sets of skis to call it a quiver, yet I'm still actively debating which new kit to pick up for next season.

I'm very solidly in Peloton's target market. I have _zero_ interest in anything related to pro cycling, and to be honest near zero interest in cycling. I needed to lose weight, and it helps to have someone on the other side of the screen pushing me to kick my own ass. I like being able to look at my workout history and re-do past classes to see if I can push a little harder. We got one mid-pandemic, and thanks to it I'm likely to exit the pandemic at the lowest weight and in the best shape I've been in over the last ~20 years. Cost was not a factor in our decision-making process.

I feel like our usernames covered that...

I will almost guarantee you two know each other (as someone who knows both of you from Google :)

... how high did you expect CPU core count to be at this point with close to uniform memory access costs, and what changes did you expect to enable that degree of concurrent bus access without catastrophic stalling due to contention?

In general, 9P2000 is simple and fairly elegant. If you don't need POSIX semantics (and fs metadata), it's really quite simple to implement. For a bit more cost in terms of complexity you can go for 9P2000.u which makes things POSIXy enough if you've got a single client (as you do with VM -> host). This is great if you want to provide virtual filesystems, either proxied from the host or backed by something more exotic. In addition to QEMU, the gVisor folks (https://gvisor.dev/) have made heavy use of 9P (with enhancements) for similar reasons (in addition to Google being generally populated with Plan 9 aficionados).

Then, there's 9P2000.L... This has that "let's just take the filesystem vtable and export it as a protocol" flavor. This works well enough on Linux if you're just looking to translate the calls over into syscalls. I wouldn't recommend implementing a bespoke server for it, though. You're better off just implementing things with FUSE.

For many things I would take HLS over writing SystemVerilog. My background is software, but I've written a modest amount SystemVerilog professionally in recent years. While there are specific cases where I'd still pick SystemVerilog (building an AXI peripheral with a specific address space layout or unique semantics for that address space), I'd tend to prefer HLS for anything actually dealing with complex data processing. Think packet pipelines, sorts, database joins, index processing, etc. If you were already in the mindset of thinking about streaming data through a pipeline (as you might in SystemVerilog) you can write fairly boring looking C++ with ordinary unit tests and the HLS tooling will do a reasonable job of turning it into equivalent pipelined Verilog.

Less than $200.

The board there is overkill -- if you really want something with a Zynq including an Arm Cortex A9 that can run Linux, the Arty Z7 from Digilent is another Zynq board that's essentially a one-stop shop for this: https://store.digilentinc.com/arty-z7-zynq-7000-soc-developm...

However, for something as simple as an NTP server you could easily get away with an Arty A7: https://store.digilentinc.com/arty-a7-artix-7-fpga-developme...

From there you can instantiate either Xilinx's Microblaze MCU _or_ you could stick to Arm and use one of Arm's DesignStart cores (Cortex-M1 or Cortex-M3): https://developer.arm.com/ip-products/designstart/fpga/fpga-...

Glancing around Digilent's catalog, they've got an even cheaper Zynq board: https://store.digilentinc.com/cora-z7-zynq-7000-single-core-...

I certainly knew about the connections between Go (and so very much other stuff at Google) and Plan 9. I was surprised, at the time, to see the OS itself in use in a production role. Just another example of unexpected tools getting the job done.

It's not entirely abandoned. A few years ago Brad Fitzpatrick dropped by my desk because the latest Google Compute Engine virtualization stack rollout had broken Plan 9 on GCE. This was interfering with cutting a new Go release (the official releases ran on VMs running Plan 9). I don't remember the specific breakage, but that was the day we added Plan 9 to the stable of operating systems we run in our CI regression suite against new VM releases.

I don't know if the Go releases are still done on Plan 9, but it seems likely that's still an active use of the OS.

> We rely on AWS, and now I'm thinking this is a giant mistake - what if they unilaterally decide one day that our business is politically unacceptable because a bad person says nice things about us?

You're making the argument for why this choice is not "free" for AWS. Not only do they lose a paying customer, but they cause some of their customers to debate the risks of doing business with AWS. It's precisely this balance of capital (including goodwill) that gives private companies the freedom to choose with whom they do business. Customers are the lifeblood of a business, but every one of them is also a potential liability. Every severance of a relationship with a customer carries more cost than merely the lost revenue.

So no, AWS _alone_ is not the arbiter of what actions are supposedly harmful to the fabric of democratic society -- the market for serving infrastructure is vibrant and competitive. If nobody in the market will do business with a given customer, well, that's capitalism for you.

(and just in case it wasn't clear, I applaud AWS making this choice, and hope the market at large will do the same)

This was my feeling in 2010 when I accepted an offer. It was a much bigger company by then, but upon joining it still felt stodgy and like any other big corp. I wonder if it still feels that way if you're there (I left in 2013).