This is how RSS/Atom should be implemented in every feed reader, but, to my knowledge, never is. That is—

If URL to the feed for alice.example.net can be found at <https://alice.example.net/feed.xml> and it is available via feed autodiscovery, it should not require the user to navigate to alice.example.net obtain the URL to the feed, and then paste that into the feed reader. Rather, when I'm looking at my feed reader, I should be able to merely indicate that I want the feed for alice.example.net. The feed reader's job, then, should be to do BOTH of the following:

• submit a request for / to alice.example.net and use autodiscovery within the reader implementation to obtain the URL for the feed (say, /feed.xml), AND
• store that fact in the user's profile data (that they want "the feed for alice.example.net"—merely the bare URL <https://alice.example.net/feed.xml>

(The URL obtained via autodiscovery can be cached.)

If ever the feed location changes, it should not require the user to then backtrack and go through the process again to obtain the feed URL. With the site alice.example.net itself being kept in the user data store, the feed reader can automate this process. A website could thereby update the URL for its feed whenever it wants—as often as every day, for example, or even multiple times a day—without it being a source of problems for the user.

To make it even better, the Atom and RSS specs can be extended so that if the user actually pastes in the feed URL <https://alice.example.net/feed.xml>, then that's fine—the specs should permit metadata to advertise that it is the alice.example.net feed (which can be verified by any reader), and the client can store that fact and need not bother the user about the overly-specific URL.

This is:

Bogart, Christopher, Margaret Burnett, Allen Cypher, and Christopher Scaffidi. “End-User Programming in the Wild: A Field Study of CoScripter Scripts.” In 2008 IEEE Symposium on Visual Languages and Human-Centric Computing, 39–46. Herrsching am Ammersee: IEEE, 2008. https://doi.org/10.1109/VLHCC.2008.4639056

UT says they have (some parts of) 47 and 48. I have confirmed it exists, but not totally clear on the completeness.

This can't be right. Sure, by this definition the relation "B overlaps-start R" holds, but so does "A overlaps-start R".

Missing March pages:

2, 3, 6, 7, 10, 11, 27, 28

PCL has some not listed here (or on HathiTrust):

https://search.lib.utexas.edu/permalink/01UTAU_INST/19i7hhk/alma991029833209706011

All right, so it doesn't say here, but UT's physical copies at PCL include what I think is the whole series.

HathiTrust's scans from the University of Michigan's collections go from volume 1 up to volume 13, but PCL has also vols. 14–17.

I don't see that Hathitrust or Google Books has "The Illustrated News of the World".

(Does this periodical have any relationship to "The Illustrated London News"?)

I've thought about and would really appreciate a platform like Lulu that would allow you to pay, say, 99¢ each to get printed copies of your favorite blog posts (standalone, not necessarily in book form), where a portion goes to the author. It shouldn't require creators to sign up for e.g. Patreon and consciously designate a priori what is and isn't "subscriber-only" content.

Like, "I just like this thing that you wrote, dude. Let me get a print copy and also make sure you get a few nickels for it."

NB: This is NOT the same as Ballou's Pictorial.

Okay, seemingly missing from here, and not made up by any of the other sources listed here are volumes:

• 61
• 62
• 69
• 73
• and 77(?)

Anything else? (It doesn't say exactly when it stopped in 1893. Vol 76 was December 1892.)

This scan has v19n3, which isn't an extant issue listed on HathiTrust (or Ockerbloom's Online Books page): https://catalog.hathitrust.org/Record/009793643

https://hypothes.is/users/mrcolbyrussell?q=%22open%20source%20is%20not%20enough%22

NOT mentioned here, according to http://www.philsp.com/mags/mcclures.html, there are seven phantom issues v56n6 and v57n1–v57n6 covering June through December 1926.

I'm not sure what happened to v56n5 or what's with the overlap of volume/issue numbers for v57n1–v57n4 (May through August 1924 and July through October 1926.)

It looks like based on what's here we are missing issue 58. According to http://www.philsp.com/mags/mcclures.html, this covers January through June 1927 (6 issues).

Does SQLite even need to construct an AST? It's just SQL. Can't it just emit the bytecode directly?

I've been critical of some of the projects/standards that the author has been involved with and adjacent efforts, precisely because of the tendency of proposals to violate the Principle of Least Power. We'd be a lot better off if more people in that cohort appreciated the architectural style of RSS- and Atom-based Web feeds and how much that impacts adoption.

Even if it is several years late, it's good that he's starting to ask questions like "is it possible to avoid using a database?" and contemplating things like the ramifications of trying to capture something in static HTML so that it can be relayed by a dumb server whereas in its original design it required, sometimes for a no good reason, an application server with smarts that enabled it to be an active participant.

This is one of those weird positions like, "It doesn't matter who you vote for. Just vote!" that people are sure to regret when faced with the right unforeseen counterexample.

This principle is behind good documentation, too. Too often programmers describe what their solution does and how it does it, but not why. Part of the why is just describing the problem that the solution is meant to address.

I found this really hard to read on archive.is (https://archive.is/YkIyW).

I used this snippet to reformat the article to manually float the "annotations" (pull-outs) to the margins:

```` javascript document.getElementById("CONTENT").style.width = "1720px";

([ ...document.querySelectorAll("[id^=annotation]") ]).forEach((x, i) => { if (i % 2) { x.style.left = ""; x.style.right = "-44ch"; } else { x.style.left = "-44ch"; x.style.right = ""; } }); ````

I'm amazed at the lack of thoughtfulness in the original post that this change of heart refers to. From http://rachelbythebay.com/w/2011/11/16/onfire/:

I assigned arbitrary "point values" to actions taken in the ticketing system. The exact details are lost to the sands of time, but this is an approximate idea. You'd get 16 points for logging into the ticketing system at the beginning of your shift, 8 for a public comment to the customer, 4 for an internal private comment, and 2 for changing status on a ticket. [...] The whole point of writing this was to see who was actually working and who was just being a lazy slacker. This tool made it painfully obvious [...]

This is, uh, amazingly bad. It goes on, and in a way that makes it sound like self-aware irony, but it's clear by the end that it's not parody.

The worst support experiences I've had were where it felt like this sort of pressure to conspicuously "perform" was going on behind the scenes, which was directly responsible for the shoddy experience—perfect case studies for Goodhart's Law.

The author says they've had a change of heart, so surely they've realized this, right? That's what led to the change of heart? Nope. Reading this post, it's this: "my new position on that sort of thing is: fuck them." As in, fuck them for not appreciating the value of this work and needing it to be done for them in the first place. The latter is described at length where they describe the jobs of the managers to already know these things—that is, the stuff that these metrics would say, if the data were being crunched. "Make them do their own damn jobs", the author says.

(I often see this blog appear on HN, and I've read plenty of the posts that were submitted to HN but have never exactly grokked what was so appealing about any of it. I think with this series of posts, it's a good signal that I can write it off and stop trying to "get" it, because there's nothing to get—just middling observations and, occasionally, bad conclusions.)

According to this catalog https://www.waverlybooks.com/home/list137.pdf, the June issue is the one that contains the poem "Goodbye to a Lady".

I don't see this on Hathitrust or Ockerbloom's Online Books Page.

Notable for publishing two poems by Dashiell Hammett.

https://storyoftheweek.loa.org/2023/05/the-main-death.html

This is:

Klosterman, Chuck. The Nineties: A Book. New York: Penguin Press, 2022. https://n2t.net/isbn:/9780735217959

To get a list of all the public domain scans, as of this writing:

```` javascript ([ ...document.querySelectorAll("table.auto-style21 a") ]). filter((x) => ( x.textContent.includes("19") && !x.textContent.includes("1929") && !x.textContent.includes("193") && !x.textContent.includes("194"))). map((x) => { let when = x.textContent; if (!when.includes(",")) when = when.split().reverse().join(" ") + " 01";

try {
  var result = (new Date(when)).toISOString().substr(0, ("1928-10-29T...").indexOf("T"));
} catch (ex) {
  console.log(x.textContent, when, ex);
}

if (when != x.textContent) {
  result = result.substr(0, ("1987-12-09").length - ("-09").length);
}
return result;

}) ````

This and the other new additions listed at https://onlinebooks.library.upenn.edu/new.html aren't listed on Wikisource and need to be added.

Hmm. I can't find anything about this! It would be nice if we had a scan.

(Since philsp.com links are unstable, this is cited as "Best Stories of All Time [v1 #6, November 1925]")

Whoa, are the 1920s issues really not that widely available?

It's also, you know, way more grokkable.

Or <label>, even...

Not quite! Your typechecker just isn't powerful enough.

Filter all lists to include just the ones with partial serials:

([ ...document.querySelectorAll("li") ]).filter((x) => (!!x.querySelector("img.info"))).filter((x) => (!x.textContent.trim().endsWith("(partial serial archives)"))).forEach((x) => (x.parentNode.removeChild(x)))

PCL has fairly complete archives: https://search.lib.utexas.edu/permalink/01UTAU_INST/vu6c1o/alma991031921079706011

This isn't made clear in enough sources.

This is:

Noseworthy, Theodore J., Fabrizio Di Muro, and Kyle B. Murray. “The Role of Arousal in Congruity-Based Product Evaluation.” Journal of Consumer Research 41, no. 4 (2014): 1108–26. https://doi.org/10.1086/678301

Meanwhile, Hypothesis (and Apache Annotator; related prior art) are both free software.

It's here: https://time.com/archive/6735152/the-selfish-meme/

I was wrong https://hypothes.is/a/p1dohn89Ee-QDVdC-4Z0oA.

Notice how bad, as of this writing, the typography is on this page.

Use this as an example of the weirdness of the difference in creator agency and consumer expectations about presentation of the written word when it comes to print vs Web.

I thought I was parsing this, but I don't understand it.

Ibid

Similarly, a few years ago I was running into a bunch of people saying stuff like, "Every programmer uses Stack Overflow. Everyone." Which is weird because it definitely had the feel of a sort of proactive defensiveness every time it came up, plus there's the fact that it's not true that every programmer uses Stack Overflow. At the time I kept running into this kind of thing, I had basically never used it, but not for lack of trying or any sense of superiority. Every time I'd landed there the only thing I encountered was low-quality answers and a realization that Stack Overflow just doesn't specialize in the kind of stuff that's useful to me. (In the years since, I've landed there quite a bit more than before, and I have found it useful—but almost never for actual programming...)

I don't think I've ever seen this in a technical discussion. Credentialism in the form of "X years experience with Y" or someone trying to flex other parts of their résumé (e.g. previous employers)? Definitely.

Most often, though, I just run into Ra + a fuckton of ho-hoism. This is never tinged by academic credentials, even a little bit.

I'm curious where this comes from. As the archives below indicate, Hathitrust only has volumes up to volume 24 (1898). UT PCL stops also at volume 24. Everything available seems to stop there.

I did chase down the Mott reference from the Wikipedia article that says it ran until 1900, but I don't remember seeing a volume count. I wonder if 1900 is substantiated anywhere else and whether the volume count is an independent claim or a derivative of the claim of cessation in 1900.

(Maybe the copyright records indicate two more volumes?)

This is listed at https://onlinebooks.library.upenn.edu/new.html, so I'm not sure why there are no volumes listed.

Hathitrust has part of the magazine under the title The New Monthly Magazine and Universal Register (1814–1820; missing vols 3, 5, 7, 9, and 10), but under the title The New Monthly Magazine, all these volumes are represented. There are at least some under the name The New Monthly Magazine and Literary Journal (1821–1836), and also some under the name The New Monthly Magazine and Humorist (1837–1852). It has the final volume in 1882 under the name The New Monthly.

So that means we won't see a classification/subclass with a range like AZ57482. But is there a limit on the number of digits after the decimal?

(Also, the literal reading of the statement here means that the range in EG9999.293 is invalid—because 9999.293 is greater than 9999.)

It doesn't look like this happened.

Huh? Is this a reference to, I dunno, draft briefs—or what? Legal texts are pretty static...

Nope. It's actually https://roy.gbiv.com/untangled/2008/rest-apis-must-be-hypertext-driven.

From https://hypothes.is/a/zpmaQHIYEe-a8iddRR3nQw

This is the first time I'm seeing this restriction.

UT has something at PCL published in 2001:

https://search.lib.utexas.edu/permalink/01UTAU_INST/3307f/alma991049170939706011

UT has 25 volumes at PCL:

https://search.lib.utexas.edu/permalink/01UTAU_INST/3307f/alma991030795149706011

UT has it at PCL;

https://search.lib.utexas.edu/permalink/01UTAU_INST/3307f/alma991034691669706011

UT has three volumes at PCL:

https://search.lib.utexas.edu/permalink/01UTAU_INST/3307f/alma991052376419706011

I'm reminded of the time when it was revealed about 10–15 years or so ago that when Apple switched to x86 from PowerPC, it wasn't the result of a big porting effort. They'd been maintaining portability all along—doing private builds internally that just never saw the light of day. When this came to light, the reaction was huge. People were awed.

A few years ago, when this piece of trivia was brought back to the forefront of my mind again after having not thought about it for years, I was struck by how silly that reaction was. Of course it makes sense that they'd been maintaining portability. There was nothing stupendous about this.

I think this is the one time when I saw and felt the effects of Apple's legendary reality distortion field firsthand. (In every other instance, I hadn't been close enough and so only perceived it from afar and only had other sources to trust that it was a real phenomenon.)

This still describes the way the semiconductor manufacturing world works.

This is sort of an odd remark. Even excluding servers and focusing only on traditional desktop and laptop computers, Windows' dominance is as weak now as it has been at any other point in the last 30 years.

Most desktop and laptop computers? Yeah, probably.

"Almost all"? Surely not.

This characterization is a good case study in the odd (off) conceptualization of computer UI...

It looks like the answer to my question is "yes".

So is this a matter of...? Taste? Or is there some rule that says which shall be preceded by a decimal and which shall not?

So the decimal does not indicate the presence (introduction) of a cutter.

Is the rule, then, that if the character following the decimal is a digit, then it's a decimalized classification, and if it's a non-digit (alpha?) then it's a cutter?

Sean Dudley pointed out something similar to me—even though a given Black Mask issue might be the January 19XX issue, it was probably actually published and distributed sometime in December. This means that if you have a serialization that began in 1928 that you want to us public excerpts from, if it it ended in the January 1929 issue, then there's a good chance that the whole thing is actually public domain.

UT has a bunch of volumes, including early volumes, that aren't already available and linked from this page.

https://search.lib.utexas.edu/permalink/01UTAU_INST/19i7hhk/alma991032048629706011

<https://purl.org/cite/ieee.org/Licklider, J. C. R. 1960. Man-computer symbiosis. IRE Transactions on Human Factors in Electronics. HFE-1: 4-11, (March 1).>

<https://purl.org/cite/ibm.com/Good, I. J. 1958. How much science can you have at your fingertips? IBM Journal of Research and Development. 2: 282-288, (October 4).>

<https://purl.org/cite/theatlantic.com/Bush, Vannevar. 1945. As we may think. Atlantic Monthly, (July). Pp. 101-108.>

I referred (indirectly) to this in an annotation on https://www.theatlantic.com/magazine/archive/1945/07/as-we-may-think/303881/ as "the PDF". As the first page indicates this is rather a PDF—specifically someone's PDF of the ACM's reprint from 1996 (which can be found hanging off this DOI: https://dl.acm.org/doi/10.1145/227181.227186).

The Atlantic's PDF can be found here https://cdn.theatlantic.com/media/archives/1945/07/176-1/132407932.pdf (at least for now).

This is a superficially attractive argument, but it doesn't hold up.

In practice, links to handles are tied to the domain of the organization operating the handle resolver service. The argument that the handles themselves are durable and can survive the demise of the domain and the org that controls it is not a good one; for everything that makes this true for handles, it's also true for names based on conventional/traditional URLs—

I find this extremely hard to parse.

UT has an incomplete copy of volumes 31 and 32:

https://search.lib.utexas.edu/permalink/01UTAU_INST/19i7hhk/alma991030494949706011

Hey, traveler.

Wayback Machine has a copy of the original HTML version.

Lassila has a copy of the PDF her homepage: https://www.lassila.org/publications/2001/SciAm.html

JSTOR (PDF): https://www.jstor.org/stable/26059207

It's almost that link rot is a problem that publishers should, you know, do something about...

baffling

No.

weird

Not to be confused with Truth, the British magazine published 1877–1957.

The book Organization and Success (1923) by William Armstrong Fairburn does not seem to be available online.

UT purports to have it at PCL, but I haven't verified.

https://search.lib.utexas.edu/permalink/01UTAU_INST/be14ds/alma991024040439706011

UT PCL has it. https://search.lib.utexas.edu/permalink/01UTAU_INST/3307f/alma991017140839706011

UT PCL has some of them. https://search.lib.utexas.edu/permalink/01UTAU_INST/9e1640/alma991028908279706011

UT PCL has them, mostly.

https://search.lib.utexas.edu/permalink/01UTAU_INST/19i7hhk/alma991015658839706011

UT has a copy that is a facsimile reprint with a newer (still in-copyright) introduction.

UT has a copy that is a facsimile reprint with a newer (still in-copyright) introduction.

UT has it in their microfilm deposit.

Kiczales is fond of the metaphor of implementing a spreadsheet by making each cell its own window under the native platform's windowing system.

This is:

Garlan, David, Robert Allen, and John Ockerbloom. “Architectural Mismatch or Why It’s Hard to Build Systems out of Existing Parts.” In Proceedings of the 17th International Conference on Software Engineering, 179–85. ICSE ’95. New York, NY, USA: Association for Computing Machinery, 1995. https://doi.org/10.1145/225014.225031.

This precedes Nelson on hypertext.

spurious "are" here

I can't seem to locate these issues. If I search Hathitrust or lib.utexas.edu, it just gives me Pall Mall. We know from a 1913 issue of Hearst's in which Chesterton's "The Treason of a Jingo" was (re-)published, there is some 1912 issue (apparently September) of Nash's in which the writer Sydney Brooks published "The Conquering English". (Chesterton's piece is a response to Brooks's.) Evidently, it is the September 1912 issue in which Brooks's article appears. However, at the time, Nash's and Pall Mall were still separate. According to Wikipedia, they didn't merge until 1914. And indeed, it looks like there are independent issues for September 1912 of both Nash's and Pall Mall. Viz:

• https://web.archive.org/web/20240806052511/http://www.philsp.com/homeville/FMI/k07/k07133.htm#A8
• https://web.archive.org/web/20240806052511/http://www.philsp.com/homeville/FMI/k07/k07133.htm#A8

Is this a reference to Taylor and the sort of work that the Gilbreth's were doing?

It's odd that this is considered to be an issue and something that I take to be a consequence of the times.

Does anybody worry about being able to remember the URLs of, say, their Golang imports?

Alternative take: if the HTML5 parsing algorithm (and its error handling) had been precisely defined, then HTML would have become as popular as fast (maybe even faster?) while being a lot more cross-compatible.

https://en.wikipedia.org/wiki/HTTP_referer

TBL says that browsers must be able to handle plain text (and not just that, but 80-column text). I wonder if this mandate appears anywhere else in modern standards (rather than just implemented by convention). It should.

(I am genuinely concerned about the possibility that browsers could/would remove support for plain text.)

Those are: - https://www.w3.org/History/19921103-hypertext/hypertext/Conferences/ECHT90/Authors.html#BernersLee - https://www.w3.org/History/19921103-hypertext/hypertext/Conferences/ECHT90/Authors.html#Cailliau

Derp. This is the wrong way to do default parameters.

The history endpoint is https://api.pocketcasts.com/user/history.

See https://github.com/DanEEStar/listening-history-deno/blob/main/src/pocketCasts.ts

#MeToo was about rape, other sexual assault, and harassment. Pratt may very well have been guilty of some or all of these things, but it's not suggested by anything in the preceding passage. (And it's pretty reductive/diminutive of the actual crimes and other transgressions relevant to the #MeToo label to point to that passage and have a response that is essentially, "lol #MeToo amiright?")

Is this accurate? I don't have a lot of firsthand experience with data science stuff, but usually when looking just past surface-level you find that some Python package is really a shell around some native binary core implemented in e.g. C (or Fortran?).

When I at the repos for spaCy and its assistant Thinc, GitHub's language analysis shows that it's pretty much Python. Is there something lurking in the shadows that I'm not seeing? Or does this mean that if someone cloned spaCy and Thinc and wrote it in JS, then the subset of data scientists whose work can be done with those two packages (and whatever datavis generators they use) will benefit from the faster runtime and the the elimination of figging and other setup?

It's hard to make sense of these comments in light of TBL's frequent claims that the Web is foremost about URLs. (Indeed, he starts out this piece describing the Web as a universal information space.) It can really only be reconciled if you ignore that and understand "the Web" here to mean HTML over HTTP.

(In any case, the remarks and specific examples are now pretty stale and out of date.)

And yet this article was published to Medium.

This foreword is described in the book as being "written as an article in 1997". There's a brief introduction (8 paragraphs dated December 2002), and then what follows is purportedly that same article, which begins, "The Web was designed to be a universal space of information[...]". The acknowledgements of the foreword, too, says that it "is based on a talk presented at the W3C meeting, London, December 3, 1997".

The same material, including acknowledgement, but sans the 8-paragraph introduction, is available on a webpage titled "Realising the Full Potential of the Web" on the W3C site. https://www.w3.org/1998/02/Potential.html

That's Reversible Object-Oriented Interpreters.

Just because they "could", it doesn't necessarily mean that they could, IYKWIM.

Dries almost but not quite caught that Solid/Inrupt was doomed.

Counterpoint: the judicious use of abstraction is/can be, in some instances, the solution to giving users agency and reconfigurability.

Software that has to be torn down is the result of software built upon bad abstractions. Abstractions are not ipso fact bad. They just need to be chosen on the criteria of whether or not they solve a problem.

This is wrong. This guest editorial actually appears in Volume 6 (not Volume 5).

As nayuki notes[1], XHTML is usable, and XHTML, being XML, supports namespaces, and this is like the one place where they're perfectly suited and not clunky.

1. https://www.nayuki.io/page/practical-guide-to-xhtml

Previously: https://holzer.online/articles/easteregg-lp-style/index.html

It would be useful to track down the misleading statement that Mozilla PR released that suggested that neither party was receiving kickbacks with the new Pocket integration. The reality is that that there was money changing hands related to the decision to integrate Pocket (NB: this was pre-Pocket acquisition by Mozilla), but the statement was worded just so to merely suggest that no money was changing hands without ever explicitly stating so—the idea being no doubt that they could claim plausible deniability wrt any false statements and blame the reader/listener for misunderstanding. The problem with this is that it backfired because it was so successful that Mozilla programmers who weren't in-the-know themselves took the statement to mean exactly the thing implied, and then they took to all sorts of public fora and "refuted" people using the PR piece, only these duped employees were explicitly claiming that there wasn't anything untoward going on, rather than the way the PR statement merely implied it. Plausible deniability moot.

(I was hoping after stumbling upon this old piece that I'd see the statement here, which would allow me to trace the contamination to e.g. HN comment threads around the same time, but this isn't the statement. It's a good clue as to when, precisely, it might have been issued.)

Dear wanderer:

You're looking for http://software.rochus-keller.ch/screenshot_oberon_ide_0.5.1.png

Dear wanderer:

You're looking for http://software.rochus-keller.ch/screenshot_oberon_system_in_debugger.png

I've been saying for years that if, say, the Gnome project wrote a new terminal emulator and replaced the default desktop terminal emulator with it and they made sure that the new one had a scrollback buffer that made your call-and-response session with the machine look more like iMessage/SMS bubbles that people are more comfortable with, people would be a lot less reluctant to use it. Once you've done that, replace Bash with an even better shell (in the same vein as the overhaul I just mentioned—not merely with something as conservative as fish or zsh), and you could magnify that effect 10x.

The real problem is:

1. public perception upon showing someone a terminal window—that they immediately adopt the thought, "Uh-oh. I don't know about this computer stuff, and I shouldn't be here"

2. the fact that the default terminal emulators do evoke that feeling

3. and the fact that tech folks are okay with this (and defensive of it, even)

Worth considering Brooks's distinction of essential versus incidental complexity. It's especially worth considering the instances where the "incidental" part comes from being incidental to the fact that if it were easier, there it would make a lot of people unhappy for reasons that I call "the consultant effect".

See also: Philip Guo on "command-line bullshittery".

Isn't that problem better fit for a solution with linked data—not by an app?

It's interesting, because there's tons of software out that has exactly one user, and then it drops off sharply for x>1 and then goes back up when you get into, what, I dunno, the hundreds?

except for the part that doesn't because it runs on the server and uses the browser as a thin client

See also Raskin on: * The Woes of IDEs * Comments are more important than code

Implicit step zero

It looks like it's this:

https://ariel-miculas.github.io/How-I-got-robbed-of-my-first-kernel-contribution/

An obvious question arises: why not just not format it like that?

How many people could you trick into using a conventional CLI if the text entry and output weren't in a window that looks like a traditional terminal emulator? What if the commands were more humane (like a step up from PowerShell) and the screen looked like you were interfacing with a not-explicitly-human agent on the other end of a messaging-like app?

This doesn't allow easy creation of bookmarklets like:

javascript:(`foo <b>bar</b>`)

We need explicit detection/support for these.

To underscore this, I've also settled on characterizing this as a problem of honesty. I put it in terms of lying—i.e. people lying about the identity (URL) of their work (Web resources).

URLs should not be considered reusable/recyclable—at least for the duration of the original publishing authority's continued renewal and control of the domain where it appeared (and even then...)

That (a compilation step) is not why you get advance warning of the sort the interviewer means when you're programming in C++ and Java. It's static typechecking that's responsible for that. You can have static typechecking without also requiring compilation.

I'm curious if Bart has ever actually tried doing this exercise.

not a bad start at a rubric

Again, you don't need that...

What does "secure" mean?

What's a "data:// url"?