HNHacker News
TopNewBestAskShowJobs

guenthert

1,704 karma · joined August 12, 2015

submissionscomments
guenthert··on Asus Bike Booster
> Have you ever ridden a bike with a geared derailleur, and forgotten to shift into low gear at an intersection, or had to suddenly stop before you had a chance to shift into low gear, so it's more difficult getting started again?

Yes, it's an annoyance, not an insurmountable obstacle. If one is bothered by it, there's the hub gear, invented more than 100 years ago.

guenthert··on Why Target Common Lisp for Code Generation?
Why not? The vulnerabilities you read about are due to oversights, not inherently lacking capabilities of the language. For every program exhibiting any given of such, you'll find thousands which don't make the same mistake at the same place. After all, most reported vulnerabilities are just a short patch away from being fixed.

Now human programmers might find it more difficult to get certain things right than others, but to a LLM only quantity of examples matters, no?

guenthert··on Because It's Not Fun Enough: why languages fail
> The killer feature used to be the REPL.

The killer (hah!) feature is arguably its homoiconicity, allowing to modify the language easily within itself ("macros"). This has the obvious advantage that you can add missing features (usually) easily yourself; and exactly that causes headaches for those tasked with maintaining other people's code.

> Nowadays with LLMs writing all the code

Oh, you came all the way from the future and that's what you bring us?

guenthert··on an ambiguity in C89 which will never be fixed
"because the problematic behavior was removed from the standard over 27 years ago, we'll never have closure on what the intended interpretation is."

Obsolete standard turns out to be insufficiently strict, news at 11.

guenthert··on Tail-call optimization in C is relatively recent (2025)
> Even SBCL doesn't do TCO at all times. Compiling at (debug 3) means no TCO.

Presumably one intends to debug the code, when setting (debug 3). Then it'll be helpful to see the stack, no?

> Another related footgun is deep recursion of other kinds, for example when recursively traversing down lists. For long lists it's easy to exceed the stack size limit. The common idiom is to recur on list elements, but iterate or map to go along a list.

Not going to argue with seasoned lispers here, but IMHO recursive code makes most sense when accessing recursive data structures.

guenthert··on Tail-call optimization in C is relatively recent (2025)
Only if they are using an insufficiently smart compiler. SBCL handles TCO just fine, as do a number of other implementations, see : https://0branch.com/notes/tco-cl.html
guenthert··on My server is a phone now
Hmmh. My Samsung Galaxy S4 doesn't boot w/o battery. Not sure, whether early in the boot process the presence of a battery is tested or whether there just isn't enough juice.
guenthert··on Making Postgres 300x faster for analytics: batching, operator fusion, and SIMD
If it is indeed 300* faster, I'm sure more rational corporations will rethink their license policy or be left in the dust.
guenthert··on Why Erdős Problems Are Falling to AI
And it's a huge distraction. Afaik (and I'm no climate researcher), the increased albedo of greater cloud coverage due to greater vaporization because of higher temperatures is the only known negative feedback effect. Everything else, be it reduced albedo due to less ice cover, increased release of Methane due to melting permafrost, increased release of CO2 due to more forest fires, etc. only accelerates global warming. So "more CO2 = more warmer" might be hilariously simplified, but it ain't wrong.

Only those who try to maximize profits while skirting the risk of a revolt care whether global average temperature will be 1.2 or 1.8K above pre-industrial average in ten years. For the rest it's already too warm, the damage is already plainly visible.

We don't need better models to predict future warming; we dragged our feet long enough that we can now look at historic data to see where it's going.

guenthert··on Windows XP 2002 for the Itanium: Unbridled rage
I'm rather surprised that they did end. After all, Itanium is one of few architectures (also the Russian Ebrus comes to mind) which doesn't exhibit the SPECTRE vulnerability.
guenthert··on Celebrating 45 Years of Kermit with the First New C-Kermit Release in 15 Years
C-Kermit? I (barely) remember tinkering with the FORTH version on a Epson HX20 (late in its life) anno 1995. The sources seem to originate (today) at https://atariwiki.org/wiki/Wiki.jsp?page=Kermit%20Protocol%2...
guenthert··on More German than many Germans
> In fact, I like most of the rules, rules are good because they ensure order, fairness and equality.

I'm sure some of the more entrepreneurial minds will argue that many of those rules stifle innovation and progress. But also on a personal level, at least to this returning German, some of the rules seem silly at best and at worst cause some underlying sense of oppression. Why, e.g., is there a Meldegesetz/Meldeamt? I survived for 25+ years in California without such.

There are safety/freedom tradeoffs and whether they are worth it, everyone has to decide for themselves (at least Germans are granted the liberty to leave, if they don't like it). Many Germans long for a more adventurous life, but in the end most prefer to visit, rather than emigrate to, foreign shores.

guenthert··on Sharing an X11 Server Across Hosts with FamilyWild
The proxy is needed for session management (specifically suspending sessions), no?
guenthert··on Sharing an X11 Server Across Hosts with FamilyWild
> Haven't tried over the internet,

It works, the user experience however might not be great. Latency is your enemy, as are occasional drop-outs. For remote connections outside a LAN, naked X11 isn't recommended. NoMachine's NX (or the shrink wrapped X2Go) were better suited for such.

guenthert··on NetBSD 11.0
> The only filesystem that was completely unaffected by any power failure was UFS, on FreeBSD (obviously with "soft updates" enabled).

And with "completely unaffected" you mean the internal (meta-data) consistency was maintained, recent user data changes were silently lost however.

guenthert··on NetBSD 11.0
Among those qualifiers would also be 'free open source'. I'd think there are hardened operating systems which are not much, if at all, talked about in public. Perhaps, in the age of LLMs finding most security vulnerabilities, those are actually demonstrably more secure now than the well reviewed public ones.
guenthert··on The Apple Calculator Language
Meh, I prefer (reverse "stressed") myself. Say what you mean!
guenthert··on Steel Bank Common Lisp version 2.6.7
Nobody stopped anyone to come up with variants, just from calling those ANSI Common Lisp compliant. There are after all many, most of which are long dead already.
guenthert··on Steel Bank Common Lisp version 2.6.7
It is my understanding that easing bootstrap was the original motivation of the fork of CMU CL. I'm a bit confused about that, since CMU CL supports compilation to byte-code (which should make a port to a new architecture comparatively easy), a feature which was dropped in SBCL.
guenthert··on OSS-SEC: 432 Linux kernel CVEs (in less than 32 hours)
> Isn't biggest attack surface in drivers?

That would be my guess too, the current batch however is all over the place: Bluetooth, file systems, etc. (the network layer ones are probably the most interesting ones). And the severity is mixed as well, some are 'just' vulnerabilities to potential DOS attacks.

> I expect distros to split less popular kernel modules into many separate opt-in packages

In the past questionable modules were blacklisted rather than removed.

guenthert··on A Road to Lisp: Which Lisp
How about the distinction between the boolean value false, the empty list and whether an empty list ought to be true or false? ]:->
guenthert··on A Road to Lisp: Which Lisp
Not sure about that. I rather think it's the very dynamic nature of Python (e.g. at run time one can add members to classes) which limits performance gains achievable by compilation. Iirc PyPy got to some five fold performance gain against CPython.

Common Lisp, as an example, makes some deliberate flexibility (and elegance) vs. performance trade-offs. I'd think to speed Python up more than PyPy could, one would have to sacrifice some of its features (some of those might not be all that popular or even desirable).

Nah, Python has its place. If its performance isn't adequate for the problem at hand, rewrite the bottlenecks in C++ and be done. LLMs should make that easy, no?

guenthert··on Linux on the Sega 32X. Who needs hardware synchronization primitives anyway?
"Didn't have the time at that moment to file a bug report"

Because you were busy working on that so urgently needed port of Linux to the Sega 32X?

guenthert··on The social physics of conversation: Communication patterns matter
'social physics of conversation'? Were social dynamics of conversations meant?
guenthert··on The four horsemen behind Postgres outages
"significantly more expensive" and "relatively small"

Well, both are true of course. You will have to put numbers on those to make sensible decisions.

Historically, creating processes in Linux has been cheaper than in other Unix-like OS (you might find results of an old Byte benchmark) and much cheaper than in Windows. Creating a thread will be undoubtedly cheaper still, but whether that will change the user experience or is worth the cost depends on details (how large the memory footprint is, pagesize, how many files are open, how many threads are to be executed, their lifetime, how many execution units are available, etc.).

guenthert··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
As long as we're pedantic ... the subject is shared memory. Unless you specify the same, non-null, target address in the call to mmap (and the kernel happens to grant you that mapping on all calling sites), the addresses will be different; the address space is not shared (each mapping might also have different access permissions).

That distinction is important as pointers generally cannot be shared (a problem which can of course be solved with one more indirection ;-) .

guenthert··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
I think it's rather unfortunate that writing regression tests is seen as a heavy burden. Compared to determining the cause of a bug and the right strategy to fixing it in a maintainable way, writing such a test should be simple, no?

I'm not at ease regarding LLM generated code changes to a project with a (hopefully) long expected life time, but LLM generated regression tests should be less contentious. I wouldn't expect them to be maintained much; rather, if they don't perform as intended against a future build, just have them recreated.

guenthert··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
This seems to be a multi-phased project. First phase (completed) was the re-write in Rust. There doesn't seem to be a performance gain and no significant one should be expected. In a 2nd phase a new architecture is implemented which malisper claims to perform much better.

I wished the two phases would have been tackled in reverse order.

guenthert··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
Dunno. I got rather the impression that it's ambitious single-developer projects with no intention of maintenance which leverage those 'AI' code generators the most.

Who wants to contribute to an unmaintainable code base?

guenthert··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
They ought to, but are they? In https://wiki.postgresql.org/wiki/Developer_FAQ I don't see a requirement to provide a regression test for a bug fix.
← PreviousPage 2 of 34Next →