HNHacker News
TopNewBestAskShowJobs

rntz

2,169 karma · joined May 25, 2009

www.rntz.net email: http://www.google.com/recaptcha/mailhide/d?k=018XOhLYKe9NqLdmoSmt6HsQ==&c=mcxnT-9mQPQEL6wailXqNUHk3aPs4Jv1T3vTJsNNAkg=
submissionscomments
rntz··on Things I learnt the hard way in thirty years of software development
If the name is mandatory, as it is in Smalltalk, then this basically solves the issue.

If the name is optional, as it usually is in eg. Python, then there is a temptation for the writer to omit it, which means the reader won't know what "True" or "False" stands for.

Are there languages besides Smalltalk and its descendants with mandatorily-named parameters?

rntz··on Monotype launches the first redesign of Helvetica in 35 years
You can find illicit copies of almost every font online for free if you know where to look. I suspect that puts a bit of a damper on the "generic clones" market strategy, since folks who _really_ want a cheaper alternative can get one for free.

Also, font companies are big and can threaten legal action, and even the threat of legal action can put people off, even if US law makes the font letter-shapes themselves not copyrightable.

rntz··on The Essence of Datalog (2018)
miniKanren is Turing-complete; Datalog deliberately isn't.

Despite that, there are things Datalog can do easily that miniKanren can't. miniKanren can't handle negation natively, but Datalog can (as long as it's stratified, that is, a predicate can't use its own negation).

Datalog is also forward-chaining, so some things that would infinite loop in miniKanren work fine in Datalog - for example, computing the transitive closure of a cyclic graph. If someone added tabling to miniKanren that would solve this.

There's also the question of performance. Some miniKanren implementations are surprisingly fast for what it does, but I suspect they can't compete with, say, Souffle, for the kind of static analysis workloads it's aimed at. Would love to be proven wrong, though.

rntz··on The Essence of Datalog (2018)
Yes, the DOOP folks under Iannis Smaragdakis have been doing this for a while. They use souffle, I believe: https://souffle-lang.github.io/.

This is Semmle's business model (https://semmle.com/). They have a custom Datalog with OO features they call QL.

I've heard LogicBlox used to do program analysis as well, but AIUI they turned more towards business analytics and then got bought out, and the Datalog engine team all jumped ship.

The Rust compiler has been exploring implementing some of their type system (the borrow checker, I think?) in a Datalog.

On the academic front, Tamás Szabó and Sebastian Erdweg have been working on an incremental Datalog engine for static analysis; https://github.com/szabta89/IncA. I know Tamas has worked for Itemis, but I don't know whether they're using this tech.

I'm working on a functional language, Datafun, inspired by Datalog, but it's too much of a toy to do any real analysis with at the moment. All in all, Datalog is still a niche language and static analysis a niche application of it, but it seems to be on the rise.

rntz··on Humans who are not concentrating are not general intelligences
> there's no reason why someone who scores well on one dimension should also score well on the others.

Except that this is literally what happens; this correlation between seemingly unrelated cognitive tests is referred to as "the g-factor". https://en.wikipedia.org/wiki/G_factor_(psychometrics)

To be fair, this doesn't actually contradict most of the rest of what you say. But this correlation does suggest that there are some shared factors (whether innate, or developed, or both) that affect many or all kinds of "practical intelligence"; one might reasonably call these factors "general intelligence".

rntz··on Ask HN: Is it just me, or is CSS too damn hard?
> Incidentally, but very likely related, I always failed to have any glimpse into how design works. I have no idea which colors "go" with which ones, and pretty much all fonts look like the same for me.

This is not related. There is no magic or natural connection between grokking CSS and having a sense for color theory or typography. They are connected by circumstance only, in that they're all useful for web design, and in that CSS can control colors and typography in the web browser. CSS is not any more an "art" thing than Javascript or C, although it is a thing with which one can make art (as are Javascript and C).

rntz··on Ideology Impairs Sound Reasoning
As far as I can tell from skimming the paper, it fails to establish that ideology impairs reasoning more than any other kind of pre-existing belief.

There is a well-established bias, called belief bias (which the paper mentions), against accepting the logical validity of arguments whose conclusions you disbelieve. The study tested examples of this where the conclusions were political (agreed with liberal or conservative viewpoints), but AFAICT did not use a control test where the conclusions were apolitical, but the participants still agreed or disagreed with them.

A control could have established whether political arguments were more (or less, or equally) susceptible to belief bias. But they didn't use one. So the study only establishes that political arguments are susceptible to belief bias.

rntz··on Ideology Impairs Sound Reasoning
> >All drugs that are dangerous should be illegal. Marijuana is a drug that is dangerous. Therefore, Marijuana should be illegal.

> The final clause does not actually follow from the rest of the logic

The final clause follows formally from the truth of the first two clauses. What you are complaining about is that the final clause is not true. But the question being asked of the study participants is not whether the argument is true, but whether the final clause follows if the first two are true. You are exhibiting exactly the behavior the study is testing: giving a wrong answer to a question of soundness by substituting for it a question of truth.

There is, perhaps, an argument to be made that "logical reasoning" in the sense being tested - being able to tell valid from invalid syllogisms - isn't all that pertinent in the real world. Is that what you're saying, though? I find it hard to tell.

> >Judge Wilson believes that if a living thing is not a person, then one has the right to end its life. She also believes that a fetus is a person. Therefore, Judge Wilson concludes that no one has the right to end the life of a fetus.

There is a logical problem with this argument, independent of any of the issues you bring up. Consider it divorced of context:

  W believes: if an X is not Y, then it is a Z.

  W believes: this particular X is Y.

  W therefore believes: this particular X is not Z.
(Key: W = Wilson, X = living thing, Y = a person, Z = may-end-life.)

The problem is that just because (X & not Y --> Z), doesn't mean (X & Y --> not Z). In other words, thinking it's okay to end non-person's lives doesn't imply thinking it's _not_ okay to end person's lives. Concretely, for example, one might think killing in war is justified.

rntz··on Implicit model of other people’s visual attention as an invisible beam
> That is, asserting dominance via staring at others is something we’re born with, and for good reason: our ancestors’ survival hinged on this when being confronted with someone bigger/stronger.

This explanation is circular. Trying to stare down an enemy who doesn't already care about staring-as-dominance is a losing strategy; they'll just attack you. It doesn't explain how staring-as-dominance came about.

rntz··on Implicit model of other people’s visual attention as an invisible beam
> Basically, men looking at women harms them, and we should avert our eyes

This is not what "male gaze" means. The idea of a male gaze, broadly speaking, is that the way women are portrayed in visual media (films, painting, photography, video games) caters to a male, heterosexual audience over other audiences, even at the expense of other aspects of the art. I think Lindsay Ellis does a good job of explaining this with examples [1]; particularly striking is the way Megan Fox as Mikaela Banes in Transformers is shot to maximise her sex appeal (to a straight male audience), so that even though the script ostensibly establishes Mikaela as savvy and competent, all that most viewers remember is that she's eye candy [2].

[1] https://www.youtube.com/watch?v=oNkTeHpHj_I [2] https://www.youtube.com/watch?v=tKyrUMUervU&t=1m05s

rntz··on Courier Prime: It’s Courier, Just Better
This is actually a common problem with digitisations of metal fonts, not specific to Courier. Many (although not all; Adobe's been pretty good about avoiding this particular failing, for example) of them are too anemic for screen use at body text size, presumably (although I have no hard evidence) because of not taking account of ink bleed. Victims include Bembo (the first one, not Bembo Book, which partially fixes this), several Caslons, Sabon, Janson, (an extreme example) Simoncini Garamond, and even Computer Modern (where the problem wasn't so much a transition from metal to digital, since Computer Modern was designed digitally, but from METAFONT - which can be parameterized, allowing "blackness" to be adjusted for ink spread - to Type 1 - which isn't).

Speculating a bit, one of the problems that digitisers may have encountered is that ink bleed affects fonts differently at different sizes. At small point sizes, ink bleed is proportionally large relative to the stroke width; at large point sizes, ink bleed has negligible effect. This is actually exactly what you want for readability's sake: at small sizes, you want to avoid thin strokes, and ink bleed does exactly that. At large size, "thin" strokes aren't actually thin any more. But a perfectly scalable digital font does not capture this beneficial effect of ink spread.

rntz··on Inter UI, a typeface designed for user interfaces
It's more accurate to say they were variable fonts. The Computer Modern used by modern LaTeX distributions was converted from METAFONT to Type 1 in 1992. In the process, specific parameter values had to be chosen - in my opinion, chosen badly, so as to make modern Computer Modern very difficult to read on screen at small sizes because it's too darn light. :(

Reasonable replacements included in most LaTeX distributions include the XCharter or libertine packages for text fonts; eulervm or newtxmath for math fonts; or mathpazo for both.

rntz··on Facebook’s Secret Rulebook for Global Political Speech
From Facebook's perspective, what is the alternative? _Not_ trying to define censorship guidelines - and consequently (as we saw in Myanmar) enabling genocide?
rntz··on A Formal Apology for Metaphysics [pdf]
> It is more than formal consistency because to show a whole number solution to a^n + b^n = c^n and n > 2, one doesn't need to use a high powered formal framework like ZFC. Thus Wiles' proof of Fermat's last theorem, which uses all kinds of infinite sets and other features of ZFC, can be falsified by a simple manipulation of finite numbers.

This follows from consistency (and is one reason why consistency is important). If there were some whole number solution to a^n+b^n=c^n for n>2, then this would be provable in Peano Arithmetic (a much simpler system than ZFC, that axiomatizes natural number arithmetic). ZFC extends PA (every theorem of PA is a theorem of ZFC, interpreted correctly; this is relatively easy to show); thus, anything ZFC proves is not falsifiable in PA, otherwise we'd have a contradiction provable in ZFC (& it wouldn't be consistent).

rntz··on A Formal Apology for Metaphysics [pdf]
> Second, and more importantly, the system of proof itself is falsifiable. If a theorem says that an equation p(x,y,z) = 0 has no solutions, and a proof without mistakes is found, then it better have no solutions. If somebody comes along and demonstrates that x=43,y=23,z=809 is a solution, then the system is in trouble. This ensures that math is not a formal game of arbitrary axioms that is internally consistent but ultimately meaningless.

I don't see how what you've said is anything other than the criterion of formal consistency - that a thing and its negation are not both provable. And a system may certainly be consistent, yet not particularly useful.

It should be noted that mathematics has not always been as rigorous as people assume. Attempts to ground math in formal logic didn't really get off the ground until the late 1800s. And this isn't a trivial point; the "rules of the game" were up for debate, and often debated, before then (and, to a lesser extent, since). People disagreed about which numbers exist; people disagree about which proof methods are acceptable; and for example, calculus - a highly practical branch of mathematics - was invented before its formal justifications were found.

rntz··on A Formal Apology for Metaphysics [pdf]
> Just consider volume: it's estimated that around a quarter of a million theorems are proved every year. That's real, measurable, progress; what's the metaphysics equivalent?

I'm not sure I disagree with your overall argument, but this point here is silly. First, I'd like to know where you got that number. (I must admit it sounds rather high to me, but I'm sure it depends how you count.)

But more importantly, measuring mathematical progress by number of theorems proved is absurd. I can write a computer program to prove theorems; it will almost surely prove more than a quarter-million a day (though eventually it will have to slow down). But they will not be _useful_ theorems. Nobody will read them, nobody will use them, nobody will build on them, nobody will find them beautiful, nobody will apply them to any practical problem. "Number of theorems proven" may be measurable, but it is not "real progress".

rntz··on Show HN: A new spline
Also as someone who's never touched splines before, I found it pretty difficult to replicate. The "a" from the screenshot looks fairly smooth everywhere (except possibly the upper-right outside curve), but I found that unless I got the positioning just right, I could run into some degree of "bumpiness". This may, of course, say more about me than about Raph's new splines :P.

If I wanted to try Raph's prior work on spiro splines for comparison, is there some way to do that that's easier than learning to use Fontforge? (When I last tried Fontforge, it didn't seem to handle hi-DPI screens well...)

rntz··on It's all about typography
Medium no longer uses a custom or proprietary font for body text. They used to use Freight Text, but now they use Bitstream Charter, which is free (in several variants):

https://practicaltypography.com/charter.html https://www.samiam.org/blog/2014-07-04.html https://ctan.org/pkg/xcharter?lang=en https://software.sil.org/charis/

Charter is an excellent and very legible font. I wish more people would use it, especially in the LaTeX world, where most people stick with the to-my-eyes nearly unreadable default Computer Modern.

A few critiques of your design:

You put linebreaks in unexpected and jarring places, breaking the flow of paragraphs. For example, in the paragraph under the heading "The Hero".

You use a narrow column for body text. This is often recommended by typography nerds, but it doesn't play nicely with elements that can't be hyphenated and split across multiple lines, such as <code> spans. See eg. the paragraph under the heading "Content font". I suggest using a wider text column to make the resulting line-breaks less jarring relative to their surroundings. You might also want to enable "hyphens: auto" in your CSS, or use a JS hyphenation library (make sure to load it asynchronously so people on slow connections can still read your content while it's loading!).

And finally, there's more to typography than choosing typefaces, as I hope the above makes clear.

rntz··on Infinite procedurally-generated city with the Wave Function Collapse algorithm
If this is based on wave-function collapse, what does it do when it hits an impossible/contradictory state?

(See https://github.com/mxgmn/WaveFunctionCollapse where it mentions "It may happen that during propagation all the coefficients for a certain pixel become zero. That means that the algorithm has run into a contradiction and can not continue.")

rntz··on Against software development
> Is the author expect to convince anyone who doesn't already agree with him?

Not really. If I wanted to convince skeptics, I would have to make it much longer and more precise, and then it be a very different kind of thing. Maybe I'll do that some other day.

The connections between the sections are fairly obscure, I admit. But I think many "universal truths of software development" are instances of more general problems to do with collective decision-making and local optimisation processes, which also produce the problems the last section mentions. I don't have a solution, but in the meantime I'd like to avoid making things worse.

rntz··on Against software development
"Getting the job done" is a short-term goal that often comes at the expense of long-term well-being. For example, the endless parade of backward compatibility layers on top of backward compatibility layers. Each got the job done at the time. And each makes it just a little bit harder for future programmers to get their jobs done.

I think a sense of beauty can let us know that something is amiss with the bigger picture. It can provoke us into stepping back and ask: what is actually necessary? Which jobs really need doing, and at what cost?

Certainly, short-termism is not the only failure mode when writing software. Abstraction for abstraction's sake is another.

rntz··on Against software development
Yes, compilers have gotten much better at optimising some forms of generic code over the past few decades. This is great! But it's not like the war is won. Staged, reactive, and incremental programming, for example, are still areas of very active research. Good implementations of these let us write code at a higher-level and yet avoid the interpretative overhead of evaluating them naively. Without such implementations people just perform the optimisations by hand (although they usually don't think of it this way) - a lot of imperative UI code is like this, for example.

Who knows what the next frontier will be?

rntz··on Against software development
It is a bit of a rant, yes; I did write it to evoke emotion; and it is less clear than it could be.

> what is the connection between "Beautiful code gets rewritten" and "Refuse to work on systems that profit from digital addictions." ?

Both result from systematic perverse incentives caused by local rather than global optimisation processes. That's part of what I was trying to get at in part II. We really aren't very good at making collective decisions yet; neither narrow ones about how to build software, nor broader ones about what software to build. I don't have a solution, but in the meantime I'd like to try not to make things worse.

rntz··on Against software development
I'm sorry it comes across that way. I certainly could have been more precise, but I wrote this to express a certain feeling, and felt that I couldn't improve it without losing something essential. Maybe someday I'll write something more analytic and less emotional.

I think a lot of code is bad, and the world is a mess. But I have no idea about your code, and I'm not sure who is at fault, or even if that's a useful question to ask. I'm certainly no genius, and I definitely don't know how to fix all of it.

rntz··on Brouwer–Hilbert controversy
This is true, but it has a problem: if you prefer constructive logic to classical because it is "finer-grained", then where do you stop?

Linear logic, for example, is finer-grained than both classical and constructive logic. Both can be embedded in it (just as classical embeds into constructive), and it makes distinctions they lack. But linear logic, too, comes in classical and intuitionistic versions, and again classical linear logic embeds into intuitionistic.

And it doesn't stop there. Linear logic can be embedded in ordered logic, and ordered logic into the logic of bunched implications. For all I know there are even finer-grained logics out there.

But doing all of mathematics in the logic of bunched implications would be an impractical nightmare, I think. I tend to prefer choosing whether to work classically or constructive (or linearly, or orderedly) according to whether it illuminates a relevant part of the problem. Sometimes you care about constructivity; sometimes you don't. If you don't, then don't waste your time trying to do the proof constructively.

The trouble, of course, is that you don't always know whether working constructively will provide illumination until you try it! :)

rntz··on Helm: Personal Email Server
You say Helm solves or will solve all these problems.

But if I don't actually need to be able to reach the box in order to use my email - if it will work even when the box is offline, powered-down, stolen, or destroyed - then what, exactly, is the point of the $500 box itself?

It's not privacy, it seems. If it were private, how could it work when the box is offline?

I'm not opposed to a "pay for your email instead of it being ad-and-surveillance-supported" model. I just don't see the point of paying $500 for a box that you're telling me the service will function just fine without.

rntz··on Helm: Personal Email Server
You say Helm literally does this. But what the grandparent was asking for wasn't just backups. They suggest (as I see it) an open standard for backups, ubiquitously implemented so it's easy to switch your provider and easy to set up new devices.

Does Helm use one of those? I suspect not, because I don't know of any such ubiquitously implemented open standard for automatic encrypted backups.

So: is Helm going to put in the effort to define such an open standard and push for it to become ubiquitous?

Even if Helm implements automatic encrypted backups, it's still contributing to a fragmented world siloed into incompatible apps and platforms. Making your siloed service better than others won't fix that problem.

rntz··on Hidden Tribes: A Study of America's Polarized Landscape [pdf]
I agree it would be nice to have a clearer picture of what anti-PC folks think of themselves as being against (and what pro-PC folk think of themselves as being for).

However, it's also useful to know people's feelings towards a phrase _without_ defining it, to understand what they feel about the term as _they_ interpret it. This gives a sense of the political atmosphere surrounding a term, which can be just as important as what the term actually means.

This isn't how things would work in some "ideal", perfectly rational world; but that's politics for you.

rntz··on The Scots Language
I think it's fair to debate whether the difference between Scots and English is between two languages or two dialects, but Scots is definitely neither a pidgin nor a creole.

Pidgins are highly simplified proto-languages that emerge to smooth communication between people with no languages in common. Scots is not highly simplified - it's about as complex as English; they share most but not all of their grammar.

Creoles are full languages which arise _suddenly_, generally within a few generations, from a mixture of two or more parent languages. Pidgins often develop into creoles when children grow up speaking them. Scots and English certainly bear influences from many language families, but these influences occurred over many centuries, not a few generations.

Scots does vary geographically, having dialects of its own (just like English), but this isn't unusual.

rntz··on The Prolog Story (2010)
On rereading, it seems like what may be happening is simply that the problem isn't very large (10-100,000 rows) and doesn't need to be solved very fast (a few minutes), and so the standard Prolog search strategy may have worked just fine. A good example of avoiding unnecessary over-engineering.

Without knowing the details of the problem it's hard to be sure, of course. Combinatorial explosion can make even searches involving a rather small number of choices take impractically long to finish. But that depends on the structure of the search problem.

Edit: The author says in the comments on the original post that he used SWI Prolog, which I believe doesn't do anything special in its search strategy, although it does have support for CLP.

← PreviousPage 2 of 15Next →