The first: Good search results.
Yeah everyone knows that right. But young people today might have forgotten the second reason we liked Google search back then:
The Second: A clean page with a single logo and search field. It loaded quick. There were no banners everywhere, no bs.
HackerNews has that feel. It is clean, information dense, and does what it needs to do.
News use to be my default homepage but surprisingly MSN gives me a much better product with more then 4 stories on the page on 32" monitors.
I miss the Google of the “to organize the world's information and make it universally accessible and useful” era.
Google Plus could have been fine if it wasn't rammed down user's throats. But no, every product had to integrate with it, had to look like it .... because ... Steve Jobs said so ? Then came Vic Gundotra with all the high craftsmanship of political empire building maneuvering and little else. Probably broke the original internal culture if it had survived till then (not sure).
I'll check out MSN.
Old Google News clone.
I'd love to see your citation on that claim.
Disclaimer: I have no opinion on this topic. I am merely speculating the potential source of info.
Another easy way to see the difference is to load each site up in a text-only browser like links. You'll find that DDG has the cleanest interface there, too.
I still use Startpage most of the time due to the generally better search results, but DDG is my backup and sometimes finds things Startpage (i.e. Google) doesn't.
Google was all about search at the beginning. I still went to Yahoo and MSN for news and information. But I think some time after mid 2000, a lot of folks began to shift from going to portal to get information, to typing keywords "news" in Google search. Somehow we are hooked to typing keywords. When Google finally released news.google.com now users could have a quick navigation of current events. This behavior is manifested in the era of social media. So many people are now getting news from logging onto facebook / twitter. To verify, as a smart reader I would search on Google, hoping to find a full version from reputable news sites.
In some countries/culture, portal is still preferred. e.g. Yahoo Japan being one. There are still some values of Yahoo.com; I still go there if I am looking up finance news or some pop-culture entertainment news (no other new sites do better than Yahoo on entertainment news broadcast).
PS. It's also the reason Google now is increasingly less special (but they do try to keep it simple when they can, subtly).
Duckduckgo is (for me) the new google in terms of speed, but the search results are not the same quality.
I'm not sure if the reason behind this is because of possible weak phone connections or that the hurricane is a good way to promote the text only site, but either way, huge shoutout to CNN for going forward with this themselves.
I've thought for a while about scraping news sites to just show their text, or something like classifying articles based on their subjects from different ones. On this front, they did it first.
All the major sites switched to something like this when the internet slowed to a halt. I assume they've always had a minimal version ready in case something like it happens again.
It's not text only but it's considerably more minimal than most news websites, loads almost instantly, doesn't have auto play content, and is considered a reputable source.
The BBC used to be similar but their international website is awful now (slow to load, more adverts, and less emphasis on actual news).
TinyTinyRSS and Fever make it a pretty good experience too.
NPR set up a text site in late 2001. A developer reworked the text site in 2005. We have made very few changes since then.
The site is a set of a few PHP scripts querying our MySQL CMS database directly with no caching. The HTML and JS were aimed at providing a decent mobile experience in 2005. That included a lot of hacks and workarounds for extremely defunct platforms.
The newsroom asked us to make some improvements. The main features:
- display more news stories following editorial order rather than reverse chron,
- remove the obnoxious interstitial "read more" view for stories,
- put the text site behind Akamai like our other web properties,
- configure HTTPS on the text site (still in progress and won't be the default due to TLS overhead)
It should be a lot faster and more pleasant to use.
Thank you to everyone who made lots of noise about CNN's text offering. There is a small contingent of developers at NPR who love the text site to the point that they've created replacements as personal projects. They were very excited to improve it for the public!
Read More ... ( 19213 bytes )
Unrelated, seems like I can't listen to podcasts. There's an <audio> tag (which doesn't use data unless you click on play if you set preload="none" ), so that's an interesting choice.
$ curl http://thin.npr.org/t.php?tid=1001
Total wall clock time: 2.7s
$ ping thin.npr.org
round-trip min/avg/max/stddev = 76.407/78.359/86.985/2.442 ms
PHP has warts, but I find it unlikely that it would be the reason for a 2.7 second page load.
* To overcome government censorship and surveillance.
* To stop internet providers from injecting ads and tracking scripts.
But the site isn't being censored? Also https won't stop the government from knowing that you connected to those servers. I agree that we want to avoid censorship and surveillance in general, but it really doesn't seem relevant here.
>To stop internet providers from injecting ads and tracking scripts.
Is your ISP actually doing that to you right now? Or is that just hypothetical?
I was on a Southwest flight earlier this week which did exactly this, using HTTP injection to display an overlay on every HTTP page. It's certainly useful to provide flight information (or Amber alerts, weather information, billing alerts), but it's Just Wrong™ to violate the integrity of a communication to do so. Perhaps there should be some standard protocol for ISPs to send messages to clients, permitting the connected OS to determine how to display them?
How do you know? Without https, a MITM-attack might already be in place, and you wouldn't even notice.
Yes there are multiple instances where my isp was injecting stuff. More frequently at public wifi spots.
Now I just need an extension that rewrites all CNN urls to this site.
0.0.0.0 ht1.cdn.turner.com # Autoplay video
0.0.0.0 ht2.cdn.turner.com # Autoplay video
0.0.0.0 ht3.cdn.turner.com # Autoplay video
0.0.0.0 ht4.cdn.turner.com # Autoplay video
0.0.0.0 ht5.cdn.turner.com # Autoplay video
0.0.0.0 ht6.cdn.turner.com # Autoplay video
0.0.0.0 ht7.cdn.turner.com # Autoplay video
0.0.0.0 ht8.cdn.turner.com # Autoplay video
0.0.0.0 ht9.cdn.turner.com # Autoplay video
0.0.0.0 a.teads.tv # Autoplay video
0.0.0.0 t.teads.tv # Autoplay video
0.0.0.0 cdn.teads.tv # Autoplay video
works for me at the moment. You never know when the blighters will change the format though.
Search for "Autoplay policy" and change it to "Document user activation is required"
I haven't test it on other sites but in Youtube works fine.
I'll connect it to my sound system if I want to hear something from the computer, which is rarely.
* base64-js (why not window.atob/btoa?)
* react-hot-loader (should not be in a production build)
* ...a bunch of other smaller modules
There's only about 10 KB of non-library application code. Note that I ignore gzip when evaluating this sort of stuff, since that many bytes of code still need to be parsed, no matter how much it compresses.
At least it's still better than cnn.com. http://www.webpagetest.org/result/170910_C8_e78305788a19b0fb...
This website contacted 65 IPs in 7 countries across 35 domains to perform 314 HTTP transactions. [...]
In total, 4 MB of data was transfered, which is 12 MB uncompressed. It took 2.753 seconds to load this page.
Compare this to the now-SSR cnn.io: https://urlscan.io/result/7d13efac-ffaa-4ac2-a091-8611e39b20... (59kB transfer)
Edit: Someone scanned the lite-version before they switched to SSR: https://urlscan.io/result/2a9690eb-3992-4712-a657-c3be959d51...
btoa("\ud83d\ude0b") -> Uncaught DOMException: Failed to execute 'btoa' on 'Window': The string to be encoded contains characters outside of the Latin1 range.
(Edit: apparently HN doesn't either; changed to \u...)
For what? This is a non-interactive text-only website. It shouldn't need anything beside HTML and CSS.
I've got NoScript on, and now the pages load perfectly. I think I may return to being a CNN reader now. If only they could figure out how to monetise this.
So even this "plain text" site still transfers ~98 % bloat and just ~2 % content.
its more around 90% JS, 10% content
After the first page load (with scripts cached) its around 87% content (unfortunately they don't seem to be caching CSS)
* Generate a HTML/DOM based on a template and data.
* Lets you efficiently update that DOM based on updates to the data and/or model (this is it's killer feature).
But if you're just going to do the first thing (generate a HTML/DOM), you don't need most of React's feature set, and you certainly don't need to push heavy JS onto the client.
That's literally the entire value prop of the website. The author isn't obligated to use your personal list of approved web tools.
Seriously, I hate this trend to bloat up websites with the rage of a thousand suns. There is absolutely no value, but it increases bandwith usage, load times and CPU/RAM usage. News websites having "fancy, modern" websites actually make me think less of them.
Fuck this web-bloat shit. The web is broken.
Some publishers are seeking to create a modern version of Life magazine. What do imagine that would look like? How about National Geographic?
I love, for example, what the teams at The New York Times are doing to enhance stories with creative uses of the modern web.
The web is not broken, it’s simply revealing that this is a better way to consume breaking news.
This is exactly why the Drudge Report has stayed relevant for 20 years.
How about a chrome extension where you can put domains on a time-based blacklist for offenses against your sensibilities.
So if you go to CNN and they autoplay an ad, you can blacklist them for 1 week - Should you forget or carelessly click one of their links, this pops an interstitial page that reminds you of their offenses and gives you the opportunity to rescind your request for a page view.
Additionally there is an opt-in for you to report the offenses to a central service for public scrutiny/shaming... thoughts?
Web infrastructure is not broken, it's just being abused. That abuse leads to some pretty incredible technology though.
Nobody needs to watch 24/7 news
I read monthlys now. Only sometimes I can't resist checking the daily (or instant) news.
...but they sure know what data rates are. Not that a website is the root of their data problems of course
An example to show that text CNN's are a thing: https://github.com/dennybritz/cnn-text-classification-tf
To add to your trend, here is an article  about "Texture Classification and Segmentation by Cellular Neural Networks Using Genetic Learning" , texture not text though as they are more vision related, though it would be interesting to see by what methods they could be applied to text.
It achieves 95% accuracy without many tricks, but the newsgroups themselves are quite different so it's not the hardest problem.
Hacker News thread: https://news.ycombinator.com/item?id=15212155
People asked for hundreds of open tabs, they got it.
So while it helps improve readability, it doesn't cut down the time to load, the amount of data downloaded, and probably helps improve the battery life only marginally (this is debatable, depending on the amount of active JS, auto-play videos and also at what time offset t the user chooses reader mode after the page loads).
I have in the past noticed that reader mode will not be available of i have killed the JS on a site that enable their print format.
They do have RSS however .
>In Hurricane #Irma’s path with a weak phone connection? Stay up to date with the text-only version of our website http://lite.cnn.io
Why couldn't they just put it on lite.cnn.com?
If you try to look up the whois for cnn.io, you only get:
Registrar: CSC Corporate Domains, Inc.
Registrar IANA ID: 299
Registrar Abuse Contact Email:
Registrar Abuse Contact Phone:
Name Server: NS-24.AWSDNS-03.COM
Name Server: NS-630.AWSDNS-14.NET
Name Server: NS-1845.AWSDNS-38.CO.UK
Name Server: NS-1268.AWSDNS-30.ORG
lite.cnn.io. 18 IN CNAME turner-tls.map.fastly.net.
turner-tls.map.fastly.net. 20 IN A 126.96.36.199
www.cnn.com. 39 IN CNAME turner-tls.map.fastly.net.
turner-tls.map.fastly.net. 18 IN A 188.8.131.52
Also, this registrar doesn't seem to provide _any_ whois contact information, is that allowed by ICAN?
Registering through CSC is difficult if you aren't a BigCo.
Very fast, but still analytics on this lite site is bothering to see!
Yeah it'd be nice if they used server-side analytics instead, but cost/benefit.
In an age when personal assistants are gaining populating, maybe sparse text-centred interfaces are set to gain more popularity?
Well, at least they tried.
There's also going directly to the AP(https://apnews.com/) for straight news articles, sans editorial, if that what you want.
If any other news org is tuning in, please do this!
I personally very much like the approach of my preferred newspaper that offers full, mobile, text only editions of its web site.
This is awesome. It's quick, to the point, efficient, and allows me to get what I need, then it gets the fuck out of my way.
I'm even willing to put money into this. Is that possible?
The only improvement I could ask for is about 100 bytes of CSS to limit the column width, change it to a serifed font, and increase the font size.
It's one of my few side projects I use every day.
> We provide an alternative startpage for news, free from left-wing bias.
font-family: 'Open Sans', Helvetica;
margin: 0 auto;
also -- compare to cnn.com from 1999: https://web.archive.org/web/20000817204102/http://www2.cnn.c...
The regular CNN site sometimes starts auto-playing videos without asking first.
Most were RSS feeds, since those used to exist. The rest were hand-written parsers, which were highly temperamental.