Is there a 'Moore's Law' for web pages? | Hacker News Reader

Is there a 'Moore's Law' for web pages?

blog.jgc.org

38 pointsjgrahamc15 comments

The HTTPArchive has been tracking Page weight (and many other stats) for thousands of websites for a few years now:

As any website performance optimizer will tell you, increasing broadband bandwidth won't really help at all with this "inflation". The biggest components of page speed is usually the number of round trip times, which only reducing latency can help with.

Interesting site - but it looks like it has only been around since 2010.

http://httparchive.org/trends.php?s=All&minlabel=Nov+15+...

Small amount of time relatively speaking (the Guardian example was from 96 through 2011). So the general trend for a bunch of sites (on httparchive.org) looks more-or-less linear, but that's probably because only a limited amount of time (2 years) is in view.

> The biggest components of page speed is usually the number of round trip times, which only reducing latency can help with.

You can also reduce the number of round trip times by using something like SPDY.

True. Only latency can reduce a single round trip time but reducing the number of serial requests also helps. In addition to SPDY, concatenating or inlining scripts / image sprites and upping the number of parallel connections helps.

I think is related to bandwidth and processing power, internet connection speed increases and if the page can be loaded in some short time like seconds then is okay and more content will be added.

Some years ago video on the web was a crazy idea now is something normal.

My guess is that eventually (a lot of) web sites will become more like games, where the highest quality experience is achieved by maximally utilizing the amount of resources available.

So in that sense we could expect many websites' bandwidth usage to follow Nielsen's law [http://www.useit.com/alertbox/980405.html]

Interesting premise - but might be a bit more convincing with a sample size larger than one (The Guardian).

Broadband is only getting roughly 10% faster per year, so it doesn't really make sense that this would be a universal law. At the very least it's unlikely to continue over the next 5 - 10 years since broadband saturation is already pretty good in the US.

I'd love to see the same calculation for all of the bytes needed to render a collection of web pages.

http://yahoo.com

http://cnn.com

http://slashdot.org

https://www.google.com/search?q=pancakes

Does anybody happen to have collected periodic snapshots of these pages on http://www.freezepage.com?

I looked around on the Wayback Machine a bit but it seems like images are not always stored.

A bit of a tangent, but if I don't ask now, I never will - has anyone applied the internet wayback machine to google maps - so a kind of "time machine" meets street view, and you see what your neighbourhood looked like 6 years ago before that condo blocked your scenic view? Would be a neat idea. If anyone's come across a solution, please share.

I believe Google Earth somewhat does it!

http://earth.google.com/outreach/tutorial_time.html

JGC, I was trying a similar computation for the Startup Framework that I'm building using Wolfram Alpha but it seems that this is available only for premium users? Am I correct, or did you just use their regular computational engine for the curve fitting?

I was just using regular Wolfram Alpha.

I thought they might be talking about the density of articles on the page growing exponentially.

The Guardian has one of the worst layouts of any webpage I regularly visit.

Reply on news.ycombinator.com