HNHacker News
TopNewBestAskShowJobs

eadmund

4,539 karma · joined February 3, 2012

submissionscomments
eadmund··on Long live Xorg, I mean Xlibre
> I understand that change can be challenging

It’s less about change and more about outright breakage. Wayland does not support me.

> You have a choice: acknowledge that Wayland is faster, more user-friendly, and more secure, or remain tied to technologies from the 1990s.

Wayland may be faster, it’s certainly more secure (in the same sense that a system embedded in a ton of concrete and dropped to the bottom of the ocean is more secure), but I’m not convinced that it’s more user friendly. It’s certainly not friendly to me, since it does not allow me to run my window manager of choice.

I’m not convinced that the technologies of the 1990s were necessarily all that bad, either. Yeah, there were some assumptions which turned out not to be the case in reality. And yeah, there is a ton of cruft. And yeah, no-one would design X11 the way it is today. But, you know what? X11, unlike Wayland, works for me.

I would love to use Wayland, honestly. That’s not a lie. But it doesn’t work. And from what I can see it doesn’t want to work.

eadmund··on The lethal trifecta for AI agents
… are ‘private data, untrusted content, and external communication.’
eadmund··on Snorting the AGI with Claude Code
> code - not pseudocode, simplified code, draft code, but actual code of a program - is one of the worst possible representations to be thinking and working in.

It depends on the language. In my experience, well-written Lisp with judicious macros can come close to fitting the way I think of a problem. But some language with tons of boilerplate? No, not at all.

eadmund··on Changes to the Kubernetes Slack
I wonder why they are going with Discord rather than Mattermost. One would think that they would have learned a lesson about relying on third parties.
eadmund··on The drawbridges come up: the dream of a interconnected context ecosystem is over
At the end of the day, servers and software engineers cost money. One way to pay for things is ads, but ads are hostile to integrations (because there is no good way to guarantee ads will be shown) — I believe this is why Twitter and Reddit killed their third-party clients. But there are alternate ways to pay for things, e.g. subscriptions. The good news here is that the sorts of things one pays for are IMHO more likely to be the sorts of things worth MCPing together. Using MCP to post to Reddit or Twitter? Low value, to oneself and to society. Using MCP to work with one’s AWS account? Higher value.

Incidentally, why do the article’s links all use strikethrough rather than underlines? Is this a deliberate style choice, or some Chrome/Firefox/Safari incompatibility? It’s pretty ugly.

eadmund··on The Art of Lisp and Writing (2003)
> Lisp is an incredible experience to write and experiment in, unlike any other language I've ever worked in.

I think that the experienced of Smalltalk is very similar in some ways, despite the languages themselves being very different. At the end of the day I prefer Lisp, in large part because I feel that it interoperates with the rest of my system better (I could be wrong — no doubt expert Smalltalkers have figure out solutions to all the issues which annoyed me once upon a time).

They’re both roads not taken. It’s sad, and I think that our industry would be in a much better position had we not wasted decades in the C trap. But history unfolded as it did. Maybe it had to.

eadmund··on GNOME and Red Hat Linux eleven years ago (2009)
That brings back memories! As does running the desktop from one’s .xinitrc:

    panel &
    background-properties --init &
    keyboard-properties --init &
    mouse-properties --init &
    fvwm2 -f .fvwm2rc.gnome
Honestly, I think that there was a lot to love about that straightforward, discoverable way of doing things.
eadmund··on Builder.ai did not "fake AI with 700 engineers"
> ChatGPT uses a combination of characters to fingerprint its output.

Does it, or are you inferring a fingerprint based on your own experiences?

> Most humans type straight quotes instead of using shortcuts like Alt+0146 or Option + Shift + ]

Well, I use X11, and ’ is an easy Compose ' >.

Most humans do all sorts of foolish stuff. Modern computers are not typewriters; we have the full panoply of Unicode at our literal fingertips!

eadmund··on Builder.ai did not "fake AI with 700 engineers"
> I’ve always used em-dashes (mild typography nerd) but recently have been considering stopping, for exactly this reason. They always flew under the radar a little, but I’d always notice when others used them, so it’s a shame and I’ll miss them.

Please don’t stop using them. Keep on doing the right thing, no matter what!

(I love em dashes too!)

eadmund··on Congratulations on creating the one billionth repository on GitHub
Just use bignums. Seriously, a decent type system can use smaller representations for efficiency and not accidentally break the world.

That way, one can represent the numbers found in cryptography as … numbers, instead of opaque Base64- or (God help one) ASN.1-encoded byte sequences.

Machine words are an efficiency hack.

eadmund··on Seedance 1.0
> There will be an abundance of talent that is finally able to create their vision rather than nepotism their way into one of 500 limited annual roles of autonomy.

Is it possible that part of what creates quality is having to collaborate with others? Would Star Wars have been as good as it was without Marcia Lucas?

eadmund··on My Mac contacted 63 different Apple owned domains in an hour, while not is use
> Is there mobile push technology which is actually fundamentally push, all the way down to the transport layer? Like open socket, listening for incoming packets only, no notifications-> no traffic?

That’s the end-to-end principle. Each host on the Internet is fully capable of listening on a socket and doing whatever its owner wants it to do.

The issue is when firewalls prevent incoming traffic, and when NAT prevents a host from even being on the internet. There’s not really a good reason for NAT with IPv6, but there are some good reasons for firewalls. They mostly boil down to human imperfection. The developers of one’s OS and software are imperfect, so the fact that a laptop sitting in Dallas can be probed by other computers in Frankfurt or Maseru thousands of times a second is an issue: a single bug will make one’s computer, and all its data, vulnerable. And users are imperfect, too. One might misconfigure one’s computer, and accidentally expose a service to the world.

There could be some approaches to mitigate these issues, but we’re probably stuck with firewalls forever. Which is really kind of sad.

eadmund··on Will the Fear of Being Confused for AI Mean That We Will Now Write Differently?
Ditto. I’ve been using them for literally decades (because the old Macs made them so easy to type). And with X11 and xkb they’re even easier!

I do worry that some of my comments have been downvoted by kids blindly downvoting anything with an em dash, but at the end of the day I prefer good style (by my lights) to imaginary Internet points.

eadmund··on The “Frankfurt Kitchen”
> A “planned order” was to replace the “senseless chaos” of the world.

The trouble with this is the implicit assumption that planned order is better. If there’s a plan, then there must be a planner, who is after all only human. He’ll have his own biases, his own emotions. And he won’t have all the information he needs. He might plan a kitchen which works really, really well for a family of four, but will it work for a bachelor? Will it work for Orthodox Jews, who need separate dairy and meat ovens, sinks, countertops, dishes and utensils?

Will it have room for innovations such as toaster ovens and smart pressure cookers (e.g. the famous Instant Pot)? What about when culture changes and show kitchens become a thing?

On the other hand, the so-called ‘senseless chaos’ of the world means treating each particular installation as its own thing.

> The notion of a regularly constructed reality corresponds to the principles of functionalism and rationality conditioned by industrial production processes. Both objects and people were to conform to these principles.

There’s another problem: demanding that people conform to the principles of industrial production processes. Every one is different in his own way — there is no average man. And it’s a short, terrible step from a system which demands that people conform to one which destroys those who don’t.

eadmund··on The “Frankfurt Kitchen”
I’m pretty sure that he meant ‘æsthetic.’
eadmund··on Tell HN: Help restore the tax deduction for software dev in the US (Section 174)
> Taxes reduces taxes?

Yes. It’s a second-order effect. Imagine if there were a 100% tax: the government would probably get no taxes, because there would be no economy.

> So are you saying that 0% rate taxes would capture the most tax?

No. There’s a sweet spot. Everyone argues about where it is, but obviously 0% and 100% tax rates would both be problems.

eadmund··on AI Saved My Company from a 2-Year Litigation Nightmare
If there is no attorney-client privilege with AI (I believe there is not), then could the opposing party demand discovery of all of one’s communications with the AI pertaining to the case?
eadmund··on Tell HN: Help restore the tax deduction for software dev in the US (Section 174)
I really liked the suggestion someone else posted in a discussion, that IP-encumbered assets should require amortisation of expenses over the lifetime of the IP (and of course the owner can always immediately write them off by releasing the IP instead).

But I do wonder what the effect on smaller shops would be.

eadmund··on Tell HN: Help restore the tax deduction for software dev in the US (Section 174)
> what is the corporate tax rate?

21%

> It's not 100%, so you're deducting a fraction developer's salary from your income, right, you're not saving that much on your tax bill each year. you're paying tax on the income you used to pay the developer.

Which is a problem if you don’t have the money to pay the tax.

Let’s combine your and the parent’s examples: 1 principal engineer @ $300,000/year; 3 engineers @ $200,000/year = $900,000/year. $1,000,000 in sales.

year 1: Company makes $1,000,000 and pays $900,000 to engineers for a $100,000 cash profit; it deducts $180,000 from $1,000,000 for a $820,000 paper profit, and owes $172,200 in taxes. Since $172,000 > $100,000, it has a $72,000 cash loss for the year. There is not year 2.

Or maybe it raises enough capital to have a cash cushion. A similar thing happens in year 2: it makes $1,000,000 and pays $900,000 to the engineers for a $100,000 annual cash profit, deducts $360,000 from $1,000,000 for a $640,000 paper profit and owes $134,400 in taxes, still more than the cash profit. The cumulative cash losses are now $106,400.

Once again in year 3 it makes $1,000,000 and pays $900,000 to the engineers for a $100,000 annual cash profit, deducts $540,000 from $1,000,000 for a $460,000 paper profit and owes $96,600 in taxes. Hey, it doesn’t owe more than it made in taxes! On the other hand, its cumulative cash losses are now $103,000. Three years, three million in revenue, 2.7 million in expenses but it’s in the hole by $103,000, still more than its annual profit.

In year 4 it makes $1,000,000 and pays $900,000 to the engineers for a $100,000 annual cash profit, deducts $720,000 from $1,000,000 for a $280,000 paper profit and a $58,800 tax bill. It still has a cumulative $61,800 cash loss.

In year 5 it makes $1,000,000 and pays $900,000 for a $100,000 annual cash profit, deducts $900,000 from $1,000,000 for a $100,000 paper profit and a $21,000 tax bill. Good news, the company now has a cumulative cash gain! At the end of five years and $5,000,000 in sales the capital owners have made … $17,200. The engineers made $4,500,000 and the government made $483,000.

In year 6 it makes $1,000,000 and pays nothing (this is very unrealistic, because in the real world every product requires maintenance …) for a $1,000,000 cash profit, deducts $720,000 from $1,000,000 for a $280,000 paper profit and a $58,800 tax bill. People complain that it’s only paying a 5.88% tax rate, ignoring the years of amortised losses. But hey, after $6,000,000 in sales the owners finally have $958,400. They take it as a dividend and it gets taxed at the top marginal rate, so they pay an additional $354,608 in taxes.

In the real world, of course, sales may or may not cover salaries, sales may increase or decrease from year to year, markets may change and so forth.

> I don't care what the law says, pay enough that no one can say that you're a lamprey on society, please.

That’s an impossible target. Any random person can say, unreasonably, that someone else is a ‘lamprey on society.’

eadmund··on Tell HN: Help restore the tax deduction for software dev in the US (Section 174)
> Maybe this is a dumb question, but if you only deduct part of their salary in the first year, what happens if you have a software developer for several years?

Not dumb at all! In the second year, you get to deduct ⅕th of the previous year’s salary and ⅕th of the current year’s salary; likewise, in the third year you get to deduct ⅕th of the first year’s salary, ⅕th of the second year’s salary and ⅕th of the third year’s salary.

The key thing is that in the fifth and following years, a business would deduct a fifth of each of the previous five year’s engineering payrolls. This is not great for a growing business, but it’s murder on a startup trying to grow from zero.

eadmund··on Congress passed socialized medicine and mandated health insurance in 1798 (2011)
Also in 1798, Congress passed the Sedition Act, which violated the First Amendment. The simple fact that a Congress included many Founders does not necessarily imply that its acts were all in compliance with the Constitution those same Founders wrote or concurred with years before.

Guaranteeing things even after people change their minds, and providing an agreed-upon process to change the agreement, is kind of the point of a constitution. Much like a contract between parties, it specifies the rules of the game ahead of time.

eadmund··on Canonicals Interview Process
No, it was not standard conversational English to use the third person plural to refer to a known individual man or woman. It has been used to refer to a non-specific person.
eadmund··on Disaster awaits if we don't secure IoT now
The issue is that one’s fridge or light bulb becomes a lodgment from which an adversary can launch LAN attacks (e.g. via Ethernet) on one’s more-sensitive devices, such as one’s personal computer (which contains more-or-less one’s entire life: banking details, correspondence and so forth).

I don’t really care if the IoT device is compromised; I do care if it is used to compromise my trusted devices.

eadmund··on A new generation of Tailscale access controls
That’s appalling. I really feel like a fool — I had thought the world of Tailscale but no longer.

This needs to be more widely known.

eadmund··on Using lots of little tools to aggressively reject the bots
> Like 'hey cool if I use this tool to interact with your site for this and that reason?'

They do ask: they make an HTTP request. How the server responds to that request is up to the owner. As in the article, the owner can decide to respond to that request however he likes.

I think that a big part of the issue is that software is not well-written. If you think about it, even the bots constantly requesting tarballs for git commits doesn’t have to destroy the experience of using the system for logged-in users. One can easily imagine software which prioritises handling requests for authorised users ahead of those for anonymous ones. One can easily image software which rejects incoming anonymous requests when it is saturated. But that’s hard to write, and our current networks, operating systems, languages and frameworks make that more difficult than it has to be.

eadmund··on Gurus of 90s Web Design: Zeldman, Siegel, Nielsen
Honestly, I believe that the Web would have been better had we stuck to those expectations more diligently and evolved more slowly and thoughtfully. That one can does not imply that one should.

Blue links and purple visited links were fine. And now on most sites there is no differentiation, and it’s sometimes difficult to tell what is a link, and a lot of sites don’t even bother linking. This is not an improvement!

eadmund··on Gurus of 90s Web Design: Zeldman, Siegel, Nielsen
> Useit didn't change its design over the years. By the Web 2.0 period, it was seen my most in the web design profession as being hopelessly outdated.

Oddly enough, I much prefer it to the corporate NNGroup site. And that last version reminds me a bit of HN itself. Simple, clean and usable — really simple, really clean and really usable, not mindlessly aping a trend (and getting it wrong) but intelligently setting its own trend.

I wish more sites adopted that style of design.

eadmund··on The Ingredients of a Productive Monorepo
> Keeping code in Dropbox kinda sucks even with 1 or .5 developers though. That said, yeah, a regular old git or (I assume, never used it) svn seems fine.

What you do is store the git repo in Dropbox, and developers just use it as a remote. With backups, this could actually go a reasonably long time, although I personally wouldn’t suggest it.

eadmund··on The Ingredients of a Productive Monorepo
> a small team of ~1.5 full-time engineers

Yes, with one and a half FTEs you should only have a single repo.

My experience with Bazel has been extremely bad, but I don’t think that it should necessarily be avoided completely. It may actually have some value on extremely large multi-team projects. But for less than two FTEs it seems like massive overkill.

I believe that you could do what you need with Kind (and maybe Tilt?), without Bazel.

> We have made this tool layer which means if I run `go` or `kubectl` while in our repo, it's built and provided by Bazel itself. This means that all of us are always on the same version of tools, and we never have to maintain local installations.

Go kind of does that for you already, with go.mod. Since kubectl is a Go program, you could achieve that goal the same way.

> it has been crucial to have an ex Google SRE on the team

I wonder how many additional team members y’all could afford in return for an ex-Googler’s salary expectations.

I sincerely hope that y’all find the maintenance expense of Bazel to be worth it going forward. Hopefully you will!

eadmund··on Power Failure: The downfall of General Electric
> a union, trade organization, or government entity. Defined benefit plans work well when risk is spread, proper actuarial oversight is used

Note that all three of those are subject to democratic capture, in which low-information voters enable the pension managers to make poor risk and actuarial decisions. Of course, a defined-contribution plan is also subject to low-information control, but at least the damage is confined to the one making the mistake, not to those unwillingly along for the ride.

I understand that defined-contribution plans must offer similar terms to both executives and regular employees, specifically to help ensure that they are fair to employees.

You mentioned low-income workers twice, but I don’t see any inherent advantage to defined-benefit plans or inherent disadvantage of defined-contribution plans for them. It is arguable that low-income workers are also less educated and less likely to make wise or even reasonable investment decisions when in charge of their own funds. Of course, they are also less likely to make wise decisions in the choice of their pension managers.

Another huge advantage of defined-contribution funds is that one owns them and can pass them on to one’s heirs, unlike defined-benefit plans.

There is an advantage to defined-benefit plans which also come with insurance (e.g. disability), but I suspect it would be more efficient to unbundle them.

← PreviousPage 6 of 34Next →