Live data from Hacker News

The web sucks if you have a slow connection

danluu.com

451–460 of 622 posts

Re: The web sucks if you have a slow connection

#451

Earlier quoted context omitted.

This sounds like the reason widening roads doesn't usually ease congestion. Which, really, can be summed up by my favorite Yogi Berra-ism "No one goes there nowadays, it’s too crowded."

> This sounds like the reason widening roads doesn't usually ease congestion. It usually does actually. What is happening there is that you have different demand levels at different congestion levels. If you alleviate some congestion by widening the road then demand goes up. That is only a problem if the demand without congestion is higher than what even the wider road can handle. As long as the new road can handle t…

No, that is still only the short term new equilibrium. What happens is that roads with unused capacity (or at capacity, but acceptable congestion) get busier as activity increases around those roads, because of the excess capacity/low congestion. Of course it's more complex than just that; it heavily depends on the spatial relationship with job activity centers within commuting distance, social expectations, economic characteristics and many more, but the core tenet remains - adding roads is not a long term solution for congestion, spatial planning is.

Re: The web sucks if you have a slow connection

#452
post #375

Earlier quoted context omitted.

That's technically true, but it assumes away the core, ever-present problems: - It may not be physically possible to add enough lanes to e.g. handle everyone who would ever want to commute into L.A. - Even if that road was correctly sized, it still has to dump the traffic into the next road, through the next intersection point. If you've increased the capacity of the freeway but none of smaller road networks that the…

It isn't physically impossible to use eminent domain to seize all the property around the roads and then build 32 lane roads all over Los Angeles. That is a separate question from how stupid that is in comparison to the alternative of building higher density residential housing closer to where people work and with better mass transit. But if people don't want to do that either , you have to pick your poison. And ther…

The Wired article posted below has a pretty good rebuttal on those ideas.

Re: The web sucks if you have a slow connection

#453

Earlier quoted context omitted.

Funny anecdote about the freeway system of southern California. When they were initially planning the system in 1930s, 40s, they were planning to have the system in use for next 100 years. So they built over sized roads (like 10 lane freeway, without having to stop for traffic lights, that go THROUGH center of a major city). When the system proved so car friendly, more and more people moved in and bought cars. Within…

Braess' Paradox: https://en.wikipedia.org/wiki/Braess'_paradox

This isn't an example of Braess' paradox.

Re: The web sucks if you have a slow connection

#454
post #428

Earlier quoted context omitted.

I've tried using Google products from Africa (Ethiopia ... last time this January), and generally, it is right out unusable. JS-heavy apps like GMail will never load properly at all. This while the connection in itself is not THAT bad. I use to use a 3G/4G mobile connection and it generally works excellent, with pretty quick load times, for everything else than javascript-heavy web apps. I have a hard time understand…

In Ethiopia's case, it's not so much the connection speed in Addis. There's a great deal of interference from the national Deep Packet Inspection filters that leads to timed-out requests, reset TCP connections, etc. JS-heavy apps make a lot of requests to background servers and should one of those requests fail, apps will hang. It's quite frustrating and I would often load pages with the console open to see which req…

>and should one of those requests fail, apps will hang

That also happens on 'good' connections, when some crappy isp router drops the packet without any icmp. The request fails only after a tcp timeout, which is large enough to be noticed. I cannot understand why asynchronous js requests do not involve smart adaptive human-oriented timeouts and why this problem is still not solved in general. TCP timeouts are simply insane nowadays.

Re: The web sucks if you have a slow connection

#456

Earlier quoted context omitted.

Appears to be mostly lack of image optimization (and he loves gifs). A common issue with blogs.

Images are meticulously optimized, problem is, retina is expensive in file size.

I'm working on a visually lossless optimization tool [1] and could reduce the images a bit further from 20.8MB to 16.11MB (-22.6%). But you're right, hidpi images is the main cost factor, adding an srcset polyfill would be a good measure.

[1] http://getoptimage.com

Re: The web sucks if you have a slow connection

#457
post #122

Earlier quoted context omitted.

I don't think the point is that everything should be human parse-able. But most things on the web are not complex applications. Dan Luu isn't trying to use Google Maps over HSCSD, he's trying to read hypertext blogs and Twitter. Do you seriously think Twitter qualifies as a "complex application"?

For a long time, you had to load 2+ mb of data to see a 140 character tweet. Twitter actually recently fixed this; the tweet text is available in the title now, so will be available very early in page load. Not only does this make twitter usable on dialup again (it was effectively unusable ever since they switched from a simple html page to a massive "application" that you have to re-download every time they deploy),…

I love that they are keen on holding compatibility with archaic SMS but not so much with slower connections.

Re: The web sucks if you have a slow connection

#458

Earlier quoted context omitted.

> Please, please, if your site requires AJAX to work at all, then retry failed AJAX queries. Anyone here have information on the most reliable heuristics to do retries? Or information on the implementations used by say Gmail or Facebook?

I have seen https://www.wikiwand.com/en/Exponential_backoff used pretty regularly in many places.

I don't see how that is relevant.

1. There is extra connection information, or information can be sampled. E.g. query to see if anything is responding.

2. Our user just wants to get action ASAP. Not necessary to be a good citizen, our user just wants it to work.

3. Heuristics depend on what works in practice. HTTP/S is a comp layered protocol so it is hard to know what is right.

4. Connection conditions are extremely varied, mobile connection type, overseas location, ISP, IPv6, proxies, VPNs, etc all affect the connection parameters so finding a reasonable heuristic is hard.

5. Sampling connection information is difficult, because when it fails you also fall to log it.

Re: The web sucks if you have a slow connection

#459
post #143

Earlier quoted context omitted.

This was the baseline experience everywhere in the 90s: people would just do something else while they waited for things to download over dialup. Clients for things like email, Usenet, browsers, etc. commonly had batch modes where you could queue large downloads so you could basically see what's new, select a bunch of large things, and then let it download while you got a cup of coffee / dinner / slept.

There was a time when Netflix was young, and I had slow internet, that if you queued up a movie and then paused it, it would continue to load. So you'd pick a movie, queue it up (literally) and then go make snacks and get situated. When the bar looked long enough you'd start watching. Then if it stalled (which it would do like clockwork every evening around 8 pm) you'd take an intermission. By the time I got stuck wi…

Yes, back when things actually buffered properly on the internet. The way YouTube buffers now: not actually loading more than a few seconds of the video ahead of where I'm watching, even when it KNOWS my internet is spotty, really frustrates me.

Re: The web sucks if you have a slow connection

#460

Earlier quoted context omitted.

It's about lag, not bandwidth.

I remember astronaut Alex Gertz somewhere saying the VNC was also for security reasons. Keep in mind that most infrastructure on the ISS was installed in the middle 00s and that the Thinkpads were possible running Windows XP and IE 6 then.

I certainly bloody hope they're not running Microsoft software on the International Space Station.
Post reply on HN