Hacker News Reader

Comment by rl3 | Hacker News Reader

Google has had terrible search quality as of late. I'm finding myself using DuckDuckGo more and more just to find very basic things that used to be easy with Google. This is just in the last couple months or so.

Not sure if that's the case here, but if it is I wouldn't be surprised.

View on HN

Quality has been going down in the last few years, not months, if you ask me.

I've got a simple (for me) use case for which Google completely fails every time. It's mostly for programmers/sysadmins though. Paste an error message from a daemon, kernel etc in google and see what you get. With a little luck, 95% of the results on the first page won't be related to your problem. I'd say 20% won't be related to the piece of software you're dealing with at all.

This can be sort of mitigated by selecting 'Verbatim', but for some reason you can't turn that on forever, you have to select it on every search.

Add censoring of search results, and Google loses its usefulness completely.

You know you can use quotes to get verbatim per phrase, right..?

Google doesn’t always respect quotes

I dont know too much about search engines, but I think that's a side effect of how they normalize their ngrams

Why would I have to do that? Not to mention that no, I don't think quotes are respected by google any more.

Quotes are indeed still respected by Google: https://support.google.com/websearch/answer/2466433

For example, this search without quotes doesn't mention Dennis Ritchie's NYTimes obituary on the first page of results:

https://www.google.com/search?q=became+more+interested+in+co...

Add quotes and the obituary becomes the first result:

https://www.google.com/search?q=%22became+more+interested+in...

That's not what people mean when they say "not respected by Google".

They're not saying quotes never work, they're saying sometimes you use quotes and google returns a page that doesn't include the quoted word.

I always use quotes when posting error messages into Google as I always get better results that way. Why does their tool work this way? I personally don't care, I'm just trying to use their black box of magic as best I can to get things done.

If you are searching a commandline argument, such as "gcc -g", you are probably not going to find ANYTHING useful at all. Why? Because the minus sign (-) means exclusion! Same thing applies for lots of other symbols that may appear in the output of a computer program. Using quotes largely solve the problem.

This has been going downhill as well, it works maybe half of the time now.

This doesn’t actually work. There is a drop down option under search tools that’s called something like “verbatim search”. I only discovered it recently so I haven’t had the opportunity to test it yet, but presumably it works.

Once upon a time I could find blog posts of people reviewing things, travel, software, equipment etc. Now I just find pages and pages of review aggregators. The advantage of finding a blog was didn't take more than a few clicks to figure out if the blogger and me shared thinking, lifestyles, financial means etc. Now all that is gone. We just have ridiculous aggregation.

The problem is exacerbated by search quality guidelines that need to be followed if you're going to rank for any competitive search term, especially when there are clear incentives for doing so - such as site speed and deploying ssl.

Coming next is the 'insecure site' warning that's likely going to cull many more hobby sites.

> Coming next is the 'insecure site' warning that's likely going to cull many more hobby sites.

Good. Setting up https shouldn't take more than an hour now with LetsEncrypt even if you have no clue about the process.

I can totally understand outdated webdesign, can't understand not having https.

I can, because I just converted a site to https, which had a combination of relative and absolute href and src parameters on it in a confused manner. This was combined with html embedded in table records.

It wasn't a simple search and replace task since several admin functionalities were also included in the site.

Think about all of the sites that have relevant information that aren't selling anything that would love to get things like this fixed, but just can't throw money at the problem.

Before you say oh it shouldn't cost that much. Think about where some of these sites are hosted and for how little. Some people don't even have the means to pay for such a fix, but they have important information to share.

This probably should have been a rule applied to sites that sell things and not sites that just have information.

I am all for making money, but I also feel like doing some charity work for this problem as well. I heard the deadline is the 24th. Anyone else interested in doing this collectively for a couple sites?

> I can, because I just converted a site to https, which had a combination of relative and absolute href and src parameters on it in a confused manner. This was combined with html embedded in table records.

Just add an upgrade-insecure-requests header to your webserver config, boom. No search and replacing needed.

https://www.w3.org/TR/upgrade-insecure-requests/

It's been added for exactly your usecase:

> Most notably, mixed content checking [MIX] has the potential to cause real headache for administrators tasked with moving substantial amounts of legacy content onto HTTPS. In particular, going through old content and rewriting resource URLs manually is a huge undertaking.

I mean, uh sure, I'll volunteer to move your sites to https, but I don't think giving a random dude on the internet root access to fix the webserver config is a good idea ;-)

This site also had mixed content due to it using a forward proxy with ARR in IIS. Since ARR doesn't forward https requests it is truly turning into a mess. There isn't an option to just move it to another webserver as that would be it's own undertaking with the dynamic part of the site being ColdFusion.

> I mean, uh sure, I'll volunteer to move your sites to https, but I don't think giving a random dude on the internet root access to fix the webserver config is a good idea ;-)

In this point yeah I agree you don't want to let just random dudes have root access to a site. On the other hand I run my own legitimate consulting business and if you think about it. Every time I am winning over a new client for all intents and purposes I am just a random dude. :)

You can always just put an nginx in front of it to terminate TLS there. Sounds like a legacy mess nobody intends to maintain anymore anyways. Or hell, cloudflare them. Kinda pointless since you won't have TLS to the backend server, but the easiest "solution". I maintain that's it's possible to inject a header and terminate TLS however messy the system is within an hour.

The world has changed though. Now that blogger is more likely than not an affiliate con man rather than a random enthusiast on software or travel.

Not that review aggregators are the answer, but right now I don’t know that there is one.

Problem is that there's a ton of money in the "review industry," (through referrals) it's very lucrative, so the big ones are spending a lot to rank high.

100% for years I would try ddg about once a year and end up going back to Google because it just didn't work for me. About six months ago I tried ddg again and have not gone back. Ddg has gotten better and Google has gotten worse.

Google’s search results would probably be useful to me if I happened to be the most generic, average person in the world. I’m guessing this is an explicit business decision and not inferior technology.

Recently, I searched "united states" and Google picked up the word "us" in a page footer -- "Contact us".

It's not just averaging of results. That's a problem, but the aggressive reinterpretation of terms is insane and makes it impossible to search for some things.

Probably it actually meant "Contact United States".

Google seems to prioritize affiliate network blogs that aggregate basic information on a subject.

This makes it difficult to locate information that exists on a term that isn't just surface level facts you can find by browsing Wikipedia.

If it helps, you can click Tools on the right under the search bar on the results page and click Verbatim. I've found that I tend to get better results when I'm looking for something specific.

This three click ritual that I have to do on probably 75% of my searches and 99% of my conscious searches reminds me on the 90s when I would have to inconveniently cntrl+alt+del on windows very frequently.

It’s an example of how google is getting more anti user. They obviously have data on this feature usage. They make a bit of a pain and don’t make it a user preference. I fear this means it’s one of the myriad of little features that developers put in that are eventually de supported.

I imagine the following scenario in a design meeting:

PM: SO THE NEW FEATURE IS AMAZING, WE WILL SHOW USERS THE MOST LIKELY RESULTS THAT THEY WANT TO SEE. AND THIS LEADS TO MORE AD CLICKS.

DEV: BUT WHAT ABOUT WHEN IT DOESNT EVEN HAVE WORDS FROM THE SEARCH?

PM: THATS NOT AN ISSUE, OUR ALGORITHM SHOWS THAT 80% OF OUR USERS CLICK ON THESE RESULTS

DEV: HOW WILL I SEARCH STACKOVERFLOW? HOW ABOUT I ADD AN OPTION TO JUST USE THE EXACT TEXT?

PM: WE HAVE TO FOCUS ON SCOPE, CANT WASTE TIME ON BONUS FEATURES

DEV: ITS NOT A BONUS FEATURE ITS HOW SEARCH WORKS BEFORE YOUR CHANGE. IT TAKES 30 SECONDS.

PM: WELL EXACT SEARCH MAKES US LESS AD REVENUE

DEV: ITS REALLY USEFUL

PM: OK MAKE IT HARD TO GET TO SO USERS DONT GET CONFUSED

DEV: OK, AND ILL MAKE A SETTING SO ITS NOT THAT BIG OF A DEAL

PM: NO SETTING THAT WOULD BE A COMMITMENT TO KEEP THE FEATURE

I dislike this meme of thinking that PMs are all useless and devs are the only sane people in product development.

Could you please not use allcaps for emphasis in HN comments? It's basically yelling.

This is in the site guidelines: https://news.ycombinator.com/newsguidelines.html.

I've found the quality to be dropping since they started to treat synonyms too loosely in the last few years. Google can mix two completely unrelated topics due to some word being a synonym for your search keyword... in a completely different context.

Reply on news.ycombinator.com