Posts by Michael Wojcik

Re: Ethical hackers are a myth.

Tron and, at current count, nine upvoters have not a clue about the current state of the IT security business. It's no wonder the industry is such a mess.

Re: The bubble has burst

For small N the Standard C qsort() is not going to perform well, because comparisons are expensive (call through a function pointer for each one). Generally, for small N, you don't care; but if you know N is going to be small, and performance is important anyway, then inlining some other sort – even one that's not O(lg N) in time complexity – is better.

Re: The bubble has burst

Quicksort can be (and often is) implemented iteratively rather than recursively, if there's a ceiling on N (which there is in all real-world implementations). It will still need O(lg N) auxiliary space but the coefficients are small and if lg N is always less than, say, 32, it's unlikely to be a problem.

Even with recursive Quicksort, if you're careful, you can keep your levels of recursion to O(lg N).

And even in a heavily-constrained environment, gnome sort would beat bubble sort, because it would use fewer registers.

Re: Pretty impressive

Shellsort is an interesting case because its theoretical performance is difficult to analyze.

Quicksort, of course, has bad worst-case performance in the naive implementation. There are various tricks you can use to avoid that, and there's Introsort, which detects adverse performance and switches to a different sort.

Radix sorts are even faster than comparison sorts like Quicksort under the right conditions; the problem is that those conditions don't fit most cases.

These days, state-of-the-art sorts seem to be things like the widely-used Timsort (a mergesort / insertion sort hybrid that employs a number of techniques for improving real-world performance and has the advantage of being stable, i.e. not reordering items with the same key) and Learned Sort (which builds a model of the input's distribution and uses that to approximate a radix sort).

For sorting large amounts of data ("external sorting"), there are various partition / parallel sort / merge approaches, some of which are fairly ornate (such as the clasic polyphase mergesort). There's a tradeoff between partitioning and merging: you can partition so none of the subfiles have overlapping keys, in which case merging is just concatenation; or you can partition randomly, and then during merge you compare the keys of the next record from each subfile. You could even do external Shell sorting where you use the Shell stride to do the partitioning and merging.

But even for small N, bubble sort is pretty terrible. It's not that easy to code or understand; I'd say naive mergesort is easier to write and read if the language has decent support for recursion and allocation of auxiliary space. And gnome sort is simpler than bubble sort, with the same performance characteristics. Selection sort is even easier to understand; you can describe that to an novice in two sentences. So even if you expect N to remain small, it's hard to see how bubble sort is ever the right choice – except when it's the only sort you know and you can't be bothered to do any research.

Re: Low bar....

Yes. Source code has two audiences: machines and humans. And the latter includes Future Me.

Readability is one of the most important factors in code quality for any software that's likely to be maintained, and that includes most non-trivial software.

Re: He needs to hire the Boring Co.

Kind of surprising Musk hasn't resurrected the Bureau of Reclamation's plan to buy water from British Columbia. That involved a lot of digging. And siphons. And general supervillain-level madness.

Re: They keep trying to build in deserts

Tech babies don't like snowy winters.

Re: They keep trying to build in deserts

Though in California, "prime agricultural land" often means "land where large agricultural conglomerates can bribe Reclamation into ignoring their rampant violations of the law while draining the Colorado River and pumping ancient aquifers dry".

There's no question California has a housing crisis, but it also has a water disaster which is only going to get much, much worse (unless they turn it into an energy disaster by trying to ramp up large-scale desalinization). Congress could in theory alleviate this somewhat by purging Reclamation and re-staffing it with people actually willing to enforce the law, which would greatly cut down on how much water is taken by the ag conglomerates, but in reality California has far too much political power for that to happen, and, hey, we want cheap almonds now rather than water tomorrow, right?

(Though eventually all that irrigation will ruin all the arable soil anyway. Keep flushing those salts out of the high country into your fields, kids. We've known since Babylon that eventually the dropping osmotic pressure and defloccination of soil will make it impossible for you to grow useful crops. This has happened to every irrigating culture in history, except the ones like Egypt (prior to building dams) that used natural flooding, which deposits fresh, low-salt soil. In the not so distant future California's agricultural sector is screwed, and it's their own damn fault.)

Foxconn never received those tax breaks

And OP never claimed they had, so what's your point? The text you quoted said that there were tax breaks, not that Foxconn profited from them.

Foxconn promised large developments. They did not deliver.

And, of course, Foxconn did not engage in these projects out of the goodness of Gou's heart. Some of it may have been hedging – entering into the venture to see whether it would turn out to be profitable – and some of it may have been quid pro quo, which might have been anything and everything from kickbacks to political capital (though Trump is wildly unlikely to repay such debts, and Walker's political worth is now minimal).

Re: Separation of business and state

"Military-industrial complex" is generally said to have been coined by Eisenhower in his 1961 farewell address. So 62 years; your recollection was right on the money.

Remember when Republican presidents were like Eisenhower? It'll be a long time before we see that again from the GOP. They've done an excellent job of turning the party into an authoritarian dumpster fire. (Lindsey Graham crying like a televangelist asking people to donate to Trump's legal defense fund is a splendid example: complete ideological bankruptcy, utter lack of shame. I don't know if the remaining handful of Republican politicians who still have any backbone are enough to rescue the party in their lifetime.)

Re: Just wondering

Due to the large size and dimensionality of the parameter space in contemporary LLMs, and the large context windows, it often is possible to elicit large passages of text verbatim from a source. It's not hard to find gradients in parameter space in these models to recover that sort of thing, particularly if you keep the temperature low enough to avoid introducing noise or annealing the output into some "better" (in objective-function terms) result.

Re: Just wondering

It's also worth noting the economics (sorry, Rikki) of classroom instruction and evaluation. Low-level courses tend to have large class sizes, often in the hundreds of students for some disciplines. Evaluating that many students quickly becomes very expensive; it's not feasible unless you use methods that permit quick, preferably automated evaluation. By definition, such responses will have quite low information entropy, and so LLMs will tend to perform well on them.

Also, at least in US higher education, introductory courses are asked to serve multiple competing aims. They are supposed to introduce students to a subject area so they'll be prepared for more advanced study. They're generally asked to perform as general-education courses, i.e. to help students "catch up" to the base level expected of undergraduates. Administrators want them to serve as gatekeepers, discouraging students who think they might be interested in a discipline but turn out not to be from continuing. Students use them for easy (or easier) credit hours to fulfill requirements. Departments staff them with the cheapest faculty – fixed-term, contingent, and graduate-student workers – and run many sections, to make their budgets. Students take different approaches to the courses depending on which of these aims align with their own motives, and many are content to simply squeak through.

Re: Cult of Elon

Sounds to me like Musk just made up his own interpretation rather than checking with the lawyers. There's no excuse for SpaceX getting this wrong.

It stands to reason that actually being with your colleagues in person will lead to a deeper understanding of each other

Assuming the consequent.

Care to try to make an argument that's not simply a logical fallacy?

Re: Ignorance is bliss

anyone with a brain knows...

Your argument might have been more convincing had you cited anyone with a brain.

Re: Hold on...

Autism is a whole bunch of things, which is why people refer to it as a "spectrum". I'm suspicious of the autism defense too, but without a lot more information I'm not going to comment on it.

Social-engineering attacks are often carried out using scripts, which can be obtained from various malware vendors. Social engineering is sometimes customized for a specific target, but often attackers just use generic approaches against a number of targets, knowing there's a good probability that at least one attack will succeed.

Jessica Clark's demonstration for Kevin Roose at DEFCON – you can find the video on YouTube – is a well-known example of a generic soc-eng attack. Clark's good, but even someone with poor social skills could learn to do it from resources like that video and some practice.

Or email messages, or, say, Reg forum posts, with empty subject lines. Madness!

Re: Next story: Proper use of CR, CR/LF & LF (and other whitespace) ...

That's Jon Postel's Interoperability Principle, not "Law". And "interoperability" is in the name for good reason: it's excellent advice for interoperability, and terrible advice for security.

Like most things, it's a trade-off.

Re: Its all about *efficient* communication...

While I am not a fan of synchronous media, and very much not of interruptions, I do use phone calls (well, Teams calls, because that's the particular hell my employer has chosen this week) at work. I almost never initiate them, but often enough I'm on the receiving end. Generally it's to help investigate a technical problem that resists analysis with a quick back-and-forth by email or chat, where we want to screen-share logs or a debugger session or some such thing; or to discuss some other sort of complex technical issue.

I take Snake's point that there are use cases where the exchange efficiency of conversation offers significant benefit. And since I work thousands of miles from most of my co-workers, and hundreds from even the closest, some sort of telephone-equivalent technology is the only way to do that.

Most of our discussions do happen via asynchronous text of one sort or another (email, comments in the review system, chat messages that are treated like email, and so forth), because that alleviates timezone differences, avoids interruption, and provides a durable record – and that last point is hugely important. I keep a lot of old email, going back decades, and probably at least once a week I'm running searches against it and finding some obscure piece of information that someone sent years ago.

(Voice transcription technology Ain't There Yet, particularly for technical discussions where there are a lot of terms of art, jargon, abbreviations, internal cant, and so forth. And current employer forbids them anyway, possibly for legal reasons – they're hell for discovery, because they add a lot of volume and are very noisy and thus difficult to search.)

Those were the days.

I still have my Usenet archives somewhere. I should dig them out; I'm sure they'd be both entertaining and informative. And, of course, Back in the Day you'd encounter a lot of people whom you'd recognize as online personalities or of real-world interest, for one reason or another. Computing pioneers in alt.folkore.computers; well-known technical experts in comp.*; authors you'd read in rec.arts.books. Many, I think most, of us went by our real names back when I was a frequently Usenet contributor, or at least our real-world identities were pretty well-known.

I don't want to go completely GOML here, but Usenet, for me, was a great improvement over its successors. Maybe something similar can be found on some Mastodon servers these days. I haven't mustered the energy to go looking.

Re: Wrapping at column 78

It's trivial to do in vim, and not that hard to do in vi with a simple word-wrapping program. Writing the latter is an exercise for the reader. When I wrote mine, circa 1990, I spent maybe an hour on it, and that was mostly because I did it in C rather than use a scripting language.

(And, yes, I'm including handling the >s. Remove them, wrap at the appropriate length, add them back. Super easy, barely an inconvenience. And having to do it encouraged trimming.)

Re: Wrapping at column 78

The reverse radix sort algorithm. Really quite clever. O(n) time and O(n) space.

Re: I miss AmiPro 3.0

Personally, I hate WYSIWYG anything. It's rarely accurate and a waste of resources, and it encourages fiddling with appearance rather than focusing on content and structure. It's like teaching composition with calligraphy pens.

Re: "it is increasingly difficult to develop entirely new features"

Shrug. I haven't needed a spreadsheet to do anything Lotus 1-2-3 didn't do. No doubt some people need the various bells and whistles in Excel, but I'm willing to bet that in most cases most users would do just fine without them, and the UI/UX wouldn't be such a horrible fucking nightmare.

Re: You must be new around here

Since the standard is loaded with undefined terms, no one knows.

Re: I miss ClarisWorks

And how many people use a chisel as a screwdriver?

There is a vast amount of Excel abuse in the world, true. I'm puzzled by the suggestion that this is a Good Thing. Certainly the ill effects are well documented.

Re: long-form writers...

Veritably, the Mount Sunflower of Microsoft's efforts in the land of document creation.

Word's dominance is proof enough of the mediocracy that governs software.

Re: long-form writers...

And LaTeX was still prominent for submissions to academic journals in technical fields back when I paid attention to that sort of thing, which was not so many years ago.

In the humanities Word seemed to dominate, unfortunately, which is one reason why I rarely submitted to humanities journals and collections. Just not worth the grief, whether it's done with Microsoft Word or LibreOffice.

Re: bizarre state of affairs

Hell, I don't care about Blu-ray;¹ I don't even think I have a Blu-ray player.² Nor do I care about commentary tracks and the like – I can't remember the last time I watched one.

But DVDs are still superior to streaming for a whole host of reasons. I can watch them when Internet service is down or congested. (On a laptop, I might be able to watch for a while even if the power's out, though I doubt the battery would last long.) I can find what I want to watch in seconds, rather than having to search through streaming-service catalogs. I can find something serendipitously: Oh, here's something I haven't seen in ages! Oh, I'd forgotten I have that! I can loan one to a friend. The UI and remote for the DVD players I have, though certainly crap, is an order of magnitude less crap than Every Fucking UI for the streaming-service apps on our set-top box, and the box's remote is an exercise in irritation and unusability.³ I will always have my DVDs; I'm not beholden to the whims of streaming services, or of the loathsome content companies like Discovery or whatever they're calling themselves now.⁴

Streaming is certainly point-and-drool convenient, as long as you don't care what you watch.

For that matter, I'd much rather have to get up off the couch, find a DVD, and insert it into the player if I want to watch something. At least that's a bit of physical activity.

¹The official spelling, as far as I can tell.

²I haven't bothered to confirm that my various DVD-playing devices aren't Blu-ray, because I don't care.

³Which isn't surprising, because it's an Apple device. My wife purchased it. To be fair, that's the only way we would have gotten one; I'd refuse to spend the money and build one myself, except I'd never get around to doing that, because I just don't care about TV and movies enough.

⁴I refer, of course, to the HBO Pogrom that Discovery embarked upon after the merger. If there were any moral order to the universe, Discovery execs would be frying in Hell.

Re: Not stupid IMHO

There are many things wrong with ICANN. There have been many things wrong with it in the past.

Removing it from Internet governance and substituting some vague UN blather will not improve the situation. A deeply flawed but still largely successful approach is better than one patently constructed from ignorance.

I'm not one of those people who believe the UN is useless – at the very least, it helps to have a ball pit where the toddlers who govern the world can go to pull one another's hair – but this is not something they can productively take over.