Page Weight Matters (2012) | Hacker News Reader

Page Weight Matters (2012)

blog.chriszacharias.com

82 pointsjanvdberg24 comments

The web has become a three-class system:

- Creators and tech-savvy people are able to cut out the bloat and enjoy a fast and secure web because they live in rich countries with good infrastructure

- The average user. Ironically the first group creates the bloated web to extract money from the average user. The average user in rich countries can count on the infrastructure to have at least an acceptable browsing experience.

- People in poor countries like Africa, which need to count on browsers like Opera to compress the web, to make it usable at all. Even with compression techniques everything is still a struggle, because the average size of web pages seems to increase more and more.

I had not expected to have to point this out on HN of all places, but Africa is not a country.

English isn't my first language, so I made a translation mistake there.

It's 99% countryside, forest and mountains, so it's not all that wrong, is it?

I know you probably didn't intend it but the first part of this comment comes off a bit elitist with the feigned shock of someone making a mistake on hn.

Oh, give him a break. Country has a non-state meaning of "area" as well.

This also shows how important it is to keep in mind whether or not you know about the causes of an effect or just about correlations:

> ... somehow the numbers were showing that it was taking LONGER for videos to load on Feather. ... I was just about to give up on the project, with my world view completely shattered, when my colleague discovered the answer: geography.

If the colleague hadn't had that hunch, he might really have given up and changed his world view - and we might have an "actually, studies at google show that lean pages are bad for latency" meme going around now, keeping people from fighting page bloat.

I have a general question. How hard is it to render a page on computer X but display it on computer / phone Y?

I don't want a proxy. I want a headless browser on a server somewhere and a head in my pocket. Something like Ember Fastboot, but for everything. Ideally it could reduce image size via aggressive lossy compression.

When I'm travelling around the most frustrating thing is not being able to browse the web on my phone because websites are garbling down 5mb a load and opening up a bunch of tracker connections.

Opera Mini does this exactly, as long as you’re OK with the rendering server being out of your control.

Opera Mini does this. Their servers would render the page for you and ship the result to your phone in an optimized, compressed form. It also supports basic JavaScript-based interactions, where the server would execute the JavaScript for you once you interact with an element (with a short deadline, something like 5 seconds) and send you the diff when it's done.

You could even use fairly complex, modern web-apps on old feature phones this way.

Very cool. Is it possible for me to run my own Opera Mini server?

Unfortunately no, it's closed source AFAIK. The cost of having this is that Opera will see all your browsing history and interactions.

I made a SaaS does this: https://www.prerender.cloud/

  curl https://service.prerender.cloud/screenshot/https://google.com/ > out.jpg

and a chrome extension that doesn't return a screenshot, but:

1. disables JavaScript 2. pre-renders the site (executes the JS) 3. returns HTML

https://chrome.google.com/webstore/detail/prerendercloud-dis...

https://github.com/sanfrancesco/prerendercloud-chromeextensi...

This is almost exactly what I want. I have two things that I still want thought:

1. (And this is obvious, I know, but...) I want to run your stack on my own gear.

2. I want this to work on my phone, which is, unfortunately an iPhone so browser extensions are blocked by Apple policy. I'd have to create my own browser. Yuck.

As an aside, are you caching public tweets, etc, by any chance?

2 layers of caching:

1. a configurable cache per URL that defaults to 5-minutes (can be disabled or increased up to 1 month)

2. and non-configurable http caching within chrome (cache-control headers), but since the cluster is auto-scaled, this cache naturally purges every few hours.

Browsh does something like this, but with more extreme bandwidth constraints in mind.

(2012)

Discussed in 2012: https://news.ycombinator.com/item?id=4957992

And in 2015: https://news.ycombinator.com/item?id=10393930

The first comment in the 2012 discussion looks a bit funny now:

> [...] I am pretty sure that graphing this out and we find the end of the web occurs sometime in 2018 when page designers and their bosses and engineers and marketing pukes have so larded down pages that the net runs out of available bandwidth and any page takes 4:33 to load.

A random newspaper or magazine article takes 2 to 3 minutes to load for me nowadays. Those 4 aren't very far away.

Most of the time seems to be waiting for servers to respond, what probably means the bottleneck is on ad-network CPUs.

Of course, at the normal web (outside of those "professional communicators" pages) things are much better.

> Most of the time seems to be waiting for servers to respond, what probably means the bottleneck is on ad-network CPUs.

A problem easily avoided by never making any requests to those servers in the first place. Block them at the point of ingress to your network (your router at home), on the clients (using /etc/hosts or equivalent) and by using blocking software in your browsers (uBlock et al). No ads, no problems.

It will invariably also mean no access to some sites which either detect ad blockers (...which can be defused using ad-blocker detectors, this is a bit of an arms race) or depend on resources from ad servers (rare for now). You can either make exceptions for those sites or just live without them.

It's quite disheartening to ponder solutions to this stuff and yet see that discussion 6 years ago from far more capable people.

Perhaps the vast majority of Internet consumers is actually ok with the status quo.

I don't think they're okay with it; they just can't pinpoint the problem, and resign themselves to the "fact" that using a computer is an inherently slow and unreliable experience, since it's been that way for as long as they can remember.

Older and non-technical people are just as vexed by bloated pages as anybody else, but without the language to describe the problem, they seem more likely to blame the reliability of their connection or even the age of their hardware rather than the poor design choices that web authors make. I've saved multiple people from replacing their "old" phones by showing them how to toggle javascript.

Or "viruses". This is what I usually hear from my non-tech family & friends. "Could you take a look at my computer? It's acting slow, I probably have some viruses."

Turns out it's not viruses, it's ever increasing website bloat that very quickly renders reasonable machines unusable. A good portion of it can be eliminated by installing an ad blocker, but this still makes people throw away perfectly good machines, because companies on the web don't care about user experience.

More cynically, the vast majority of Internet customers are OK with it.

Given the choice between having the consumer wait 10 more seconds for the page, vs. not knowing whether the ad company screwed them / leaving 1c on the table in data not collected & sold... they'll happily choose to waste someone else's time, every time.

Reply on news.ycombinator.com