Hacker News now runs on top of Common Lisp

Regarding the "worse is better" discussion: At least its definitely better accessibility-wise. HN is about the last well-known site that allows interacting with it, including writing comments, with plain old Lynx. I am well aware that most web devs do not care anymore these days, and they have their reasons for sure. However, its still nice to see sites that refuse to go for SPA. It makes them so much more useable for people like me (blind). A big THANK YOU to the site maintainers, its one of the last corners of the net where interesting stuff happens which is still accessible.

Hacker News is a perfect example of the &quot;Worse is better&quot; mantra applied to social engineering. I mean, slashdot had more features and functionality in the late 1990s.<p>What makes HN work is the tight focus and heavy moderation.

So, Hacker News was not rewritten in Common Lisp. Instead they reimplemented the Arc Runtime in Common Lisp.

Lisp supremacy approaches!<p>I&#x27;m reminded of definitively the most extreme writing on programming I&#x27;ve ever read, here <a href="https:&#x2F;&#x2F;llthw.common-lisp.dev&#x2F;introduction.html" rel="nofollow">https:&#x2F;&#x2F;llthw.common-lisp.dev&#x2F;introduction.html</a>, including but in no way limited to claims such as:<p>&gt; The mind is capable of unconsciously understanding the structure of the computer through the Lisp language, and as such, is able to interface with the computer as if it was an extension to its own nervous system. This is Lisp Consciousness, where programmer and computer are one and the same; they drink of each other, and drink deep; and at least as long as the Lisp Hacker is there in the flow, riding the current of pure creativity and genius with their trusty companions Emacs and SLIME, neither programmer nor computer know where one ends and the other begins. In a manner of speaking, Lispers already know machine intelligence---and it is beautiful.<p>Has any other language produced such thoughts in the minds of human beings? Maybe yes, but I don&#x27;t know of one. Maybe Forth, or Haskell, or Prolog, but I haven&#x27;t found similar writing. Please do share.

&gt; Arc was implemented on top of Racket<p>Originally on MzScheme, then later PLT Scheme. It was ported to Racket by the great kogir, IIRC.

The article makes it sounds like Dang also helps with the codebase. There must be others, but Dang is the one I&#x27;ve seen for years at this point.<p>I&#x27;ve beeing a part of many online communities as both a member and moderator. However, Hackernews is the community that I&#x27;ve been apart of for the longest and the one that brings me the most joy.<p>Dang, is there anything random people like me can do for you? Can I at least buy you a coffee or something?

&gt; [Clarc] is much faster and also will easily let HN run on multiple cores<p>This was all running on a single core??

If dang is listening: I&#x27;d like your comments on how to pull this off. Replacing the engine of a live site without some old forgotten part breaking is hard to accomplish. I rarely see this kind of thing happen without a week of frantic bug fixing and users grumbling.

Good, sbcl it&#x27;s great for CL. And now with current CLX from QuickLisp (the one with daily releases, I can&#x27;t remember it&#x27;s name) MCClim runs snappy even under Intel n270 ATom machines. Under ECL it almost runs snappy, but the performance gain it&#x27;s astronomical. From a really laggy UI to instant rendering.<p>EDIT: UltraLisp for QuickLisp.

sbcl is a workhorse. I wonder if the Racket folks didn&#x27;t consider the Arc under production workloads general purpose enough to fix. I actually don&#x27;t know of any other projects that use Racket in anger<p>I&#x27;ll always have a soft spot in my heart for Armed Bear because that JVM library ecosystem is enormous <a href="https:&#x2F;&#x2F;github.com&#x2F;armedbear&#x2F;abcl">https:&#x2F;&#x2F;github.com&#x2F;armedbear&#x2F;abcl</a>

As someone who runs a website based on the Arc code that was opened sourced... I&#x27;d love to be able to use Clarc.

Hmmm. Does that mean we&#x27;ll get dark mode now?

Are there any other popular (&gt;10k DAUs) sites that still use an esoteric, homegrown tech stack? If you have worked on them, what do you think, is it a legacy mess nobody wants to touch, or a pleasure to work with?<p>PG made an assertion once that websites (in contrast to desktop software) are free to use any stack of their choosing, as long as it can take in HTTP requests and output JSON or HTML. This intuitively seems to be true, especially so with how powerful modern machines can get, but it seems like it hasn&#x27;t increased stack diversity much.<p>The advantages of boring technology and &quot;resume-driven development&quot; seem to outweigh whatever gains you may get from using something custom.

Definitely makes more sense than Racket imo, Common Lisp is a lot more pragmatic and SBCL is like magic.

Rewrites are definitely not “always a bad idea” as Joel Spolsky once said. What they are is highly situational.<p>HN has a bunch of factors that make it amenable to a rewrite. It has gigantic scale, not a ton of complexity at a business level, and what it “is” is pretty slow moving at this point.<p>That means it’s not a great example to justify a rewrite at work :) that said the success does prove rewrites are possible. Bravo on shipping!

The success of Hacker News doesn’t come from flashy features, but from a community that consistently produces high-quality content. That said, I can’t help but wonder if there are any updates to the UI&#x2F;UX in the works, LOL.

Is it still Paul Graham&#x2F;Robert Morris working on it? Skimmed the article but did not see a ref.<p>need to check out what it adds to CL: <a href="http:&#x2F;&#x2F;arclanguage.org&#x2F;" rel="nofollow">http:&#x2F;&#x2F;arclanguage.org&#x2F;</a>

Not saying &quot;security by obscurity never works&quot; but am saying it&#x27;s a shame the defensive wall of anti-spam&#x2F;abuse depends on some secrecy, because it&#x27;s a low wall. If the concern with knowing the secret sauce is how easy it would be to defeat, then its a low wall. But, as long as it stays secret, it&#x27;s doing it&#x27;s job.<p>I&#x27;m not an infosec professional, or a competent LISP coder, I&#x27;m not in a position to say what&#x27;s better. This is just what pros in the field say to me.<p>(It&#x27;s mentioned in the article)

I assume it&#x27;s worth it to keep it in Arc and not rewrite in something more widely available, is that so?

That&#x27;s great, but I think they should improve the responsiveness. It&#x27;s still a bit wonky. You can see it on the top right corner if you narrow the screen and then widen it

&gt; anti-abuse measures that would stop working if people knew about them<p>A heavy lesson in that for other implementors of discussion-forum cum blog-comment systems.

Is this open source software that I can run my own hacker news as well?

When we will get to see the code of clarc? I hope that there is no &quot;business logic&quot; relevant to running HN in the language implementation, is there?

Hacker News is the url I use to test most fussy connections because it&#x27;s so light, and will load under even the slightest trickle of data. When I was doing research in Ghana, it was the only site I could get to reliably load for news in the field, and thus spent a month reading only HN (good luck getting the _New York Times_ to load without a gigabit connection). Appreciate how it stays — and has stayed — svelte and fast throughout the years.

&gt; Hacker News now runs on top of Common Lisp<p>&gt; there’s now an Arc-to-JS called Lilt, and an Arc-to-Common Lisp called Clarc.<p>&gt; But Clarc’s code isn’t released, although it could be done:<p>&gt; Releasing the new HN code base however wouldn’t work:<p>I&#x27;m not sure if I follow all that. If the Clarc is not released, then how does HN run on it?

I cannot believe how people are praising a centralized, heavily censored links site. Take a look at HN&#x27;s privacy policy, they sure do make money from monetizing every single thing you say and also fingerprint you all the time. We should have a decentralized link sharing site

Random question, how big is hacker news? It’s plain text so I’d imagine it’s reasonably compact?

Dang! Thanks dang!

could probably have saved themselves a lot of trouble and asked Claude to rewrite it in C++

You should have rewritten it in sh and called Sharc, or BASIC and called it BArc, or PHP and called it PHarc, or ML and called it MLarcy. Or license it under GPL-3 and call it Gnarc!

Is the bel project still alive?

Look, I like the way HN looks but there aren&#x27;t many sites that essentially look like bare html but still struggle with displaying more than 300 comments.

&gt; Much of the HN codebase consists of anti-abuse measures that would stop working if people knew about them. Unfortunately. separating out the secret parts would by now be a lot of work. The time to do it will be if and when we eventually release the alternative Arc implementations we’ve been working on.<p>Is this a case where security through obscurity is good, or bad? Legit question. I am curious to read the responses it may prompt.<p>I found this though: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=27457350">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=27457350</a><p>&gt; There are a lot of anti-abuse features, for example, that need to stay secret (yes we know, &#x27;security by obscurity&#x27; etc., but nobody knows how to secure an internet forum from abuse, so we do what we know how to do). It would be a lot of work to disentangle those features from the backbone of the code.<p>The question still stands for curiosity!

&gt; Much of the HN codebase consists of anti-abuse measures that would stop working if people knew about them.<p>We’ve all heard about how “security through obscurity” isn’t real security, but so many simple anti-abuse measures are very effective as long as their exact mechanism isn’t revealed.<p>HN’s downvote and flagging mechanisms make for quick cleanup of anything that gets through, without putting undue fatigue on the users.

&gt; Much of the HN codebase consists of anti-abuse measures that would stop working if people knew about them. Unfortunately. separating out the secret parts would by now be a lot of work<p>The business logic in encoded into the original structure, making migration to anything different effectively impossible - without some massive redesign.<p>This, I think more than any response, indicates why the philosophy of “it’s working don’t touch it” will always win and new features” requests will be rejected.<p>HN didn’t depaginate based on user desires, it was based on internal tooling making that feature available within the context of the HN overall structure.<p>HN has zero financial or structural incentive to do anything but change as little as possible. That’s why this place, unique in the internet at this point unfortunately has lasted.<p>HN is not *trying* to grow, it’s trying to do as little as possible while staying alive; so by default, it’s more coherent to maintain because its structure isn’t built for it and changing the structure would break the encoded rituals (anti-abuse measures).<p>Something to think about when you’re trying to solve for many problems like “legacy code” “scaling needs” etc… it all comes back to baseline incentives

Just a note on the grammar of this: &quot;HN runs on top of SBCL since a few months&quot;.<p>&quot;since a few months&quot; sounds wrong, it isn&#x27;t idiomatic english. Consider replacing it with:<p>&quot;HN has been running on top of SBCL for a few months now.&quot;

Hacker News has so little capability, almost any experienced developer using a modern AI Coding Agent could replicate the entire thing in a weekend, and perhaps in a single day.<p>I&#x27;m not saying it&#x27;s bad, or criticizing anyone. I mean it does what it does, and it works, and people like it. But no one should care what technology they&#x27;re using because there&#x27;s just nothing impressive going on from a technical perspective.