Hacker News Reader

Comment by Etheryte | Hacker News Reader

I feel like this article overlooks a very simple idea at the heart of what Stack Overflow is: there are only so many questions you can ask about e.g. how to write a CSS selector before all the others are duplicates. This is not necessarily a bad thing, the whole idea of Stack Overflow is to have answers to general questions, not to everyone's individual question. A lot of people, especially newcomers, don't understand this though, and then get frustrated when their question gets closed as a dupe. Closing dupes is a feature not a bug, just like all roads lead to Rome, you want your questions to build a linked list to a good canonical question with a quality answer. If the whole site is riddled with a hundred thousand questions that were caused by a typo then everyone would get drowned in the noise. In other words, as far as I'm concerned this is working as intended, Stack Overflow is a QA site, not tech support.

View on HN

> Closing dupes is a feature not a bug

Closing actual dupes? yes. Closing as "dupe" questions that sound similar to other questions, that's a massive bug, and happens if "close all dupes" is a major goal for people, over "ensure I understand the question and the question I think it's a a dupe of" (which can be quite hard and specific!), the overall context (yes, the question was answered in 2009 with "use jQuery". No, that's not the answer in 2024, but there is no process to "re-open" a question where the state of the art has changed), ...

> but there is no process to "re-open" a question where the state of the art has changed)

Of course there is. When you get closing privileges you also get re-open privileges. And, believe it or not, people do reopen questions when it's appropriate.

    the overall context (yes, the question was 
    answered in 2009 with "use jQuery". No, that's 
    not the answer in 2024

This is the crux of SO's problem. At scale, it's just an impossible task for the moderators to parse out this sort of context since the problem space (the number of potential dupes) is massive and getting larger every day.

At a minimum, for that problem to be overcome, IMO there would need to be a lot of metadata associated with each question so that moderators could make faster and more accurate judgement calls regarding the question of, "is question A a dupe of question B? are question B's answers still relevant?"

This seems like an issue where LLMs could actually be of some assistance. In my experience ChatGPT has been pretty good about understanding that sort of change over time and which solutions are deprecated. I wouldn't want LLMs making the decisions outright, but I could see them being a valuable assist for human moderators.

> Closing dupes is a feature not a bug

Yes, but it's implemented in a user-hostile manner that kills engagement and therefore retention.

A friendlier implementation imo, would be for the asker to be shown the duplicated answer(s) inline under their question, with a note that it's a dupe and a link to the original question(s). That leaves them with a vastly different experience, and regular users can still continue to see it as just closed. Kinda like shadow banning, but healthier.

> A friendlier implementation imo, would be for the asker to be shown the duplicated answer(s) inline under their question, with a note that it's a dupe and a link to the original question(s).

That would occupy a lot of space. What's wrong with the current approach of posting a link to the duplicate? Mind you, when the first user votes to close as a duplicate, the system automatically posts a comment with a text more or less like "this question already has an answer on this link http://stackoverflow.com/.... If you think this case is different please state why". Then after three votes the question gets closed.

But please notice it can still be reopened. Users with 3000+ reputations (so not only what many people percieve as sort of barons on stackoverflow) can vote to reopen. And three votes are enough to reopen a question.

>>What's wrong with the current approach of posting a link to the duplicate?

Simple. The user's problem is not solved. The whole point of StackOverflow was to help people get answers to their questions.

If you fail at this, you lose your relevance.

>>Users with 3000+ reputations (so not only what many people percieve as sort of barons on stackoverflow) can vote to reopen. And three votes are enough to reopen a question.

You don't have to under go these rituals in front of LLMs just to get your problem solved.

Even until now, StackOverflow had no competition, so it didn't matter. Now that there are alternatives. People won't put up with user hostile experience.

Instead of closing dups, I think they should go into a pile and "a link" shows you the list. The reader can then sift through them. Literal duplicates could still be closed from there. I've run into many closed dups where the original answer, while still correct, didn't click for me where the closed question got responses that gave me the Aha! moment.

SEO the hell out of the "pile pages" so the top hits go to something useful, as opposed to a closed question with everyone commenting that it's the top Google result.

This is already the case, all linked questions are shown in the right sidebar and you can easily navigate through the graph, both up and down.

"the whole idea of Stack Overflow is to have answers to general questions, not to everyone's individual question"

But why? I ran a site called WPQuestions.com for 7 years, it was where people paid between $10 to $50 to get fast answers for their WordPress questions. I called it "micro consulting" in that the consultants were answering small questions for small fees. While I get that StackOverflow wanted to be free, I wonder why it didn't commit to being more of a Helpdesk for people in need. Why have general answers, when people have such an urgent need for specific answers to their specific questions? An incredibly wasted opportunity. Even more so, with AI able to answer all of the general questions, the only avenue open to non-AI companies is answering very specific questions that are heavily context dependent.

Answering everyone's specific personal questions is geared towards beginners and doesn't add any value to others. E.g. fixing your typo doesn't help the next guy understand why their code doesn't work. Explaining the general case is timeless and valuable to all skill levels. There are plenty of sites that do what you describe, but there's a reason Stack Overflow became widely known and used instead: the QA is applicable to everyone working in that space, not just that one person who asked about their specific case.

I think StackOverflow has convincingly demonstrated that “timeless” in programming is about 5 or maybe 10 years.

And that’s because programming languages move so rapidly, whether they need to or not. For some reason I can’t quite fathom there is always a desire to shake up the status quo, that particularly as a mechanical engineer I don’t understand. I don’t want some radical new way to integrate components, let alone new different components, I just want something that’s reliable and will be usable for a long time.

Depends on your use case in terms of the time pressure you're under, but I learnt a lot from SO answers. Having to figure out how a general answer applied to a specific question taught me much more than getting a specific answer to a specific question.

Non-rhetorical question: is it possible to be both? That is: be a support/micro-consulting site and a repository for canonical knowledge and best practices?

Stack Overflow sort of flirted with this idea, I guess, with the concept of attaching "bounties" to your questions. The bounties were just Stack Overflow points, not cash. But AFAIK that wouldn't spare your question from being marked as dupe so you still couldn't use SO for support.

That would be correct if technology doesn't advance and change constantly, which is not the case. New frameworks, new technology, new versions for programming languages are constantly on the rise, so I would expect that will be a good reasons to ask more new questions on SO, but that's not the case anymore.

I think you did some "grammar typos" with the two "not the case"?

IME the problem is that the criteria for "dupe" is way too low. Like, I start with a question asking how to do X in .NET8, which is what I am trying to answer. But it's closed as a dupe of something else that answers the "same" question for .NET framework 4.5, an API that was deprecated 10 years ago. Part of this might be Microsoft's awful naming choices, but some is also Stack Overflows obsession with no dupes.

    there are only so many questions you can ask about e.g. 
    how to write a CSS selector before all the others are 
    duplicates

Right, but there's a staggering number of permutations of other factors.

Think how many ways there have been to center a freaking HTML element over the years! There are, at a minimum, at least 10-20 "correct" answers to that question over the years.

Even today, that answer varies a lot. What is your render target? A state of the art desktop browser with automatic updates turned on? Browser in some locked down corporate/government/military environment that is 5 or 10 or 20 years old? HTML embedded in email? Some combination of those?

But Stack Overflow's design lacks that kind of specificity - there's no metadata associated with questions that would reliably specify the specific environment and software/library versions being targeted.

And even when the person asking that question does give that information, the moderators don't have the time or the inclination to manually and laboriously parse out those specifics and compare them to the specifics of the other possible dupe candidates... a comparison space which grows by zillions of candidates per hour.

> Closing dupes is a feature not a bug

The problem is with uneducated mods who falsely close a question as a dupe.

Separating real dupes from possibles dupes is a hard task, but if they are not up to the task then SO needed to find people that were or change the system.

This is a good take on that admin like identifying dupes is a feature and can be grating for newcomers.

But I don't think their is a practical upper bound on good questions though, so the reduction in total questions and answers in those graphs is definitely not good.

I have always thought the scale itself is very hard to maintain. I thought the different sites had a good balance but it is hard work. Mods and superusers should get compensation.

SO wouldn't be nearly as annoying and avoided if they bothered to noindex or 301-redirect closed-duplicate questions. I mean, both of those would also be terrible, but at least it would reduce noise in all of our search results.

pretty much, the article doesn't take into account all factor. There is so more question you can ask until you end up having most question already answered.

The 'dated' user interface is irrelevant, people don't scroll it endlessly like they do on facebook. they come for question and answer, the ui is fast and well made for that purpose.

> the ui is fast and well made for that purpose.

The UX not so much though.

There's so many x.stackexchange.com estates that often overlap (devops, serverfault, linux, askubuntu etc).

The search is, and has always been, very poor to unusable. I've always just reverted to DDG or google to search stackoverflow.

Having loads of answers and questions in a clean and fast UI is good, but if no-one can actually find the right Q and A, the overall "interface" is just outdated and poor.

(Edit: related: I've always voiced the opinion that "duplicates" aren't solved with more or harder moderation, but with better search)

Reply on news.ycombinator.com