We had managers and even non-engineers check-in documentation at Google: each page had an "Edit me" button that spawned an editor in a new tab with a CL(PR) ready. It worked very well.
You can feel how you wish. Few dispute one should be free to express themselves in a way that does not impact others. You can do whatever you want in your own head. It is when your views butt against others that issues inevitably arise.
I do not get your point. Is it an issue that I have an opinion on policies that impact other people? Are you against the concept of causing issues?
I think we cause issues all the time. For example, by importing trash that we do not need. It is a part of living. But the we need to address these issues, like a tax on imports. This will inevitable lead to other issues, such as the small scale import of books not being economically viable. But the fact that it causes issues is, in my mind, not a good enough argument against it. Because all things that impact other people cause issues.
> You can't deny that Sun hardware was top tier and always ahead of its time (until Oracle bought them).
From Brian Cantrill's original HN comment in 2011(https://news.ycombinator.com/item?id=2287033): "I never hesitated to tell anyone that was listening that our x86 boxes were starting to smoke the hell out of UltraSPARC". SPARC stopped being ahead of its time way before being bought by Oracle.
>> You can't deny that Sun hardware was top tier and always ahead of its time (until Oracle bought them).
> SPARC stopped being ahead of its time way before being bought by Oracle.
You seem to be focusing strictly on performance, but there were other factors: e.g., remote management is still a bit of a mess with x86. With SPARC everything can be done on the CLI (serials consoles or SSH), with x86 quite often you're still clicking on "open KVM window" for so many things. And of course many/most vendors make you purchase an 'advanced license' to use the remote KVM functionality of the LOM system.
I currently admin several racks of Dells (and Lenovo, and SuperMicro, and previously HP). Just make sure you buy the license for KVM in case you ever need to go into single-user/rescue mode (otherwise better not mind having to drive to the DC and hooking up a crash cart).
Ask @bcantrill (the author of this article/story) about how well integrators integrate (as a hint to the answer, he co-founded Oxide Computers because he ran a cloud service with Dells/HPs/etc).
Integrators integrate very well at the level that they're operating, i.e. the single server. Rack-level integrated systems are a whole different thing, very specialised and unrelated to the old thing.
It's path dependence. If you built a Sun-syle x86 server with a serial console and inserted a random Linux installer CD, the installer would probably only output VGA and thus it wouldn't work.
> I would wager that almost no Linux installs are done by CD or displaying VGA anymore (maybe 0.01%).
When one of your Proxmox (or XCP-ng, or VMware/Nutanix/OpenStack) servers won't boot and you need go into single-user/rescue mode, then talk to me about the convenience or not of a remote KVM (versus having to drive to the DC and hook up the crash cart).
It's all very well if all you manage is VMs/instances, but some of us deal with the hardware that those VMs run on, and things break.
That 0.01% of servers is probably the layer that the other 99.99% run on (and yes, a lot of that 0.01% can also be automated, but if PXE isn't working (or you're getting strange CPU3VOLT or SLOT17ECC or BUS1RESET errors) then neither is re-imaging).
Because pretty much every BMC out there in last decade is shipping AST2400/2500/2600 which include VGA and you want to deal with cases where people do not know how to navigate serial console, especially in case of windows servers
Yeah, that was amazing. And so, I learned how to provide LILO + agetty on COM1 on every server I was admining :) So, unless there was HW fault, I could just connect to server via serial to check things out.
> Modernism is the act of using cheaper resources and more standardized/cheaper methods of production.
That's how they sold it, but it's not true. By the end of the 19th century, all the building ornamentation that you can still see in European capitals, had been stadardized and was mass-produced, while the construction techniques using brick and mortar had been further refined, yielding building designs that were cheap to construct and easy to maintain.
Modernist buildings were, at the same time, more expensive to build and less durable because of the reinforced concrete which has a much shorter lifespan than stone or brick plus mortar. The only advantage they had was that of increased height, but that was only an issue due to the restrictive zoning in Anglo countries. Without NIMBYism that pushed towards sky-scrapers, NYC could well have ended up like Paris: replacing 1-3 stories buildings with 8-10 stories, but allowing mixed commercial/residential space everywhere.
Literally had nothing to do with "NIMBYism." That's anachronistic bullshit. It had to do with the desire to remake the country into Garden Cities with lawn-based neighborhoods filling the spaces between shopping malls.
The people in the places where these neighborhoods were built didn't get a say, they were removed and new middle-class "YIMBYs" were moved in to be little lords on their little lots.
Given that models have been trained on tons of academic papers as well as internal corporate documentation, it's no surprise their output sounds a lot like an unholy fusion of both styles.
It looks nice but I can't load any comments, and I believe it's because I have Firefox's Enhanced Tracking Protection turned on to the strictest level.
I see no mention of foreign keys, or any other constraints, across shards. If, as I suspect, they're not implemented, it would still be useful but at the level of Spanner 10 years ago.
Are you one of the developers ? If so, you need to add a specific page explaining in detail the guarantees that this gives.
For example: point 08 says "Assign different tables or workloads to different shard groups" and point 02 says "Split hot shards as workloads grow". How do those interact ? Can a single table be split across multiple shards ? If so don't you need 2pc to enforce primary key constraints ?
Different people have different feelings. Bucketing into binary "good" and "not good" based on whether they have the exact same framing as you do is dehumanizing. Stop it. Respect other people.
There is space for nuance:
If you're spitting out raw LLM output and not looking at or reading it and requiring other people to use it as-is, while representing it as your own original work, you are treating your own time and attention as more valuable than that of others. You can expect a reaction based on this.
If you're blanket-dismissing someone else's work because they used one or more specific tools, regardless of how much of their own time and attention they put into it before sharing it in a place where you were able to see it -- or perhaps you're doing that dismissal without even seeing it -- you are treating their time and attention as superficial and a rounding error compared with their choice of tools. You can expect a reaction based on this.
In between, there is a _whole_ lot of room for different take and opinions, and we're collectively figuring out what's acceptable and what isn't. We'll never fully get there and that is Okay. But we can at least try to calibrate on "are we treating other people respectfully" instead of a shitty proxy for it.
One of these things was wholesale systemized actual state-run murder. The other is a range that encompasses big companies run by awful people and also a binary blob I can run on a GPU in my basement.
reply