Optimizing a Static Site | Hacker News Reader

Optimizing a Static Site

hackernoon.com

58 pointsNicoJuicy27 comments

I have a hard time understanding why it is justifiable to use 1.8mb of javascript to produce a static website. The author's site is plain text, stylized and opens a modal (with animation) when you click on a link. It looks great, but I don't see why 1.8mb worth javascript is required to produce this (of which there a good chance <5% is actual code for the site's functionality/styling).

When it comes down to it, what you should be optimizing for is efficiency of delivering information to your users. Animations don't help there in cases like this. And I doubt the javascript is helping either.

EDIT: 1.8mb figure is from the bundle.js file from the author's live website

The reason why there was such a small effort on optimization in the first place is that this isn't meant to be a revenue generating site, or one that focuses on "ultimate performance" - it's more of just a personal, online resume. Any time I needed something new, I'd just throw it in, because "it worked and looked good on my machine". If I found a cool js snippet that needed an entire library, I'd just throw it in. An argument could be made that that's part of what's wrong with modern webdev, and I wouldn't necessarily disagree, but this wasn't trying to be the fastest or most efficient site out there. I think there's a pretty distinct difference between making a scalable, well optimized webapp and just making a personal site as an online resume.

This was me documenting my updates - I just revised the article with a few more updates, and the total site size is 250kb, 99kb without images. This still isn't great, but total page speed is pretty quick (js execution takes ~70ms on my 2016 MBP).

Thanks for the feedback!

I know JavaScript is cool and I'm not saying it's not a good idea to try things out and experiment -- it is! But I personally feel an individuals resume/personal site isn't what should stand out. Make it a true static site, lose the bloated JavaScript, add some basic styling and it will do what this site is doing -- maybe do a little more as people will not have to click on things, wait, read, close, click, read, etc. Let your projects speak for your skill set, not a fancy resume.

All the best.

Yeah, it's pretty easy to get a 10x increase when you are stripping out nonsense.

JavaScript doesn't necessary have to slow down the page. I recently optimized a full blown web app with basically just JavaScript to fully load in less then 200 ms.

> JavaScript doesn't necessary have to slow down the page.

If I have to wait for it to download, it sure will slow down the page.

> [...] fully load in less then 200 ms

On how fast of a connection? It's one thing for you web app to load quickly on my university's (very nice) network, but quite another for it to work on my apartment's fritzy, bussy wifi, or at the hackerspace when 10+ other people are sharing 2MB/s.

Firefox Dev Tools has a "Responsive Mode" (see top right icons) available that offers to throttle the network speed. This might be a good way to benchmark loading times.

I picked "regular 3d" in firefox and loading time went up 10x. So it seems size does matter. But 2s is not that bad, it's better then some "native" apps.

"However, it’s only 829kb, and that includes every single non-image asset (fonts, css, all libraries and dependencies, and js)."

Only 829KB? ONLY?

This[1] is a (dynamic) on-line course for all beginner sport shooters in Poland, each page takes about 15KB. And i still sometimes want to rewrite the CSS, as it's now 3 requests. It shouldn't make a difference on HTTP/2, but that's a lame excuse.

It's not very fast (100-200ms) and maybe i should get a faster server -- there's room for improvement. But in my view, if it's static, it's supposed to be as fast as a native app on your smartphone. 1MB of bullshit js that nobody wants is not "ONLY".

Btw, maybe mine is lame and ugly. But i couldn't sleep if i made people load 1MB of fonts, javascript and CSS. It's just a shameful waste.

[1] https://patentstrzelecki.eu/ (polish)

Agreed. I tried taking this to the extreme with my personal site. It's down to ~30ms on free Netlify for first and subsequent loads, always <10kb including content, with one HTTPS request per page load. No external dependencies, reasonably readable, 100% passing in Lighthouse. The highest leverage things are usually to delete anything font related, reduce or remove media and GZIP/minify. More obscure approaches include base64ing a fake favicon to get rid of the extra call, eliminate media, eliminate all requests but the core HTML static resource, inline everything, move to a faster host with lots of CDN edges, test aggressively with slow connections and eliminate until <100ms. Mostly because it's fun, also useful though if you serve a user base in developing world.

I also moved to netlify (from a gh-pages/cloudflare setup). And while I'm a huge fan, and loved the process, I am interested that google crawler seems to have found it a lot slower.

https://i.imgur.com/RUssvAA.png

I've SSL enabled (as well as HSTS) And yes, the initial redirect seems to take a while. But equally the very first page request from a site is actually quite slow I feel. Considering it's a static page of less than 4kB, I'd expect it to be a bit snappier...

I've other reasons for trying netlify, and I'm not too bothered about the slowdown (my site has bugger all traffic tbh), but I would like to get to the bottom of it.

Could you link your personal site if it's not too personal?

I had a similar reaction. I was hoping the article would touch on less obvious things, and I even read it as 829 bytes the first time (then was confused for a second). Hard to take it seriously after that.

You're right of course. One can even create a simple single-page application without using frameworks and external libraries and fit it in 5% of the mentioned size. I built verbatim [1] in 24KB, JS, CSS, all image assets included. It is a an SPA and works pretty fast, scoring 100/100 on both desktop and mobile Google PageSpeed tests.

The issue is when trivial applications grow in complexity. Any custom approach is seldom scalable. Especially when more people are brought on to work on a project and have to get up to speed. There is a strong argument to using popular frameworks (react, angular), which comes at a cost of massive app bundles. It is not, however, reasonable to use for static sites in any case.

[1] http://verbatim.link/

Original author here - thanks for your feedback, just updated the article. I kind of settled at ~1mb because I thought it was good enough. I got it down to ~90kb (which still isn't great, but is fine for these purposes) by just stripping out font awesome.

Thanks!

Jon, I run a few sites in a similar space. I feel 1MB is fine for my customer base. If HN isn't your customer, don't worry about them. If HN is your customer, you have the hardest job in the world because none of us will ever pay you anything you can build a business off of.

Go forth and prosper amongst the people who will pay you for the value you provide. Ignore us.

That's exactly what I thought before. I was starting to think that it was shared as a kind of joke or sarcasm towards web development in general.

I'm ashamed to be calling myself a web developer... What's up with everyone including tons of code from others, code they don't even know.

"Wooah, we need to toggle a class here!" "We got jQuery for that!"

"We also need to.." "Oh, we got this for that!"

Take this page as an example. It freaking huge, and the writer seems to think he has done a good job optimizing it now.

The page is really simple, and still a 1.66 MB JS bundle is required to even begin rendering the page.

And still the page doesn't work properly if you just do something as simple as view one external link and press the back button in an attempt to get back.

I know of a JS framework that might fix it, and it's only 7.6 MB. /s

Seriously, people should write about stuff they know.

Thanks for the reply. I agree that 1.66mb is insane for what my site does, and I'm working on minifying it as much as possible. I got it down to ~90kb, which is still not amazing, but is significantly better than 1.6mb. Fonts were taking up more than 90% of that, which I just stripped.

I do disagree with the idea of including code from others being a bad practice/smell. The detriment and added load time of using a library like jQuery is well worth the functionality it provides, IMO.

Thanks for the comment and for pointing out a problem with the article, though - I've revised it!

Just a quick vote to reinforce what you did. You made something, and asked for the most absurd audience to critique it.

Keep being you. HN hated Dropbox, of all things. I am hesitant to announce my product here because I struggle with as widespread of negativity as I perceive here.

If someone isn't going to buy what you're selling, I feel they should be silent and just not buy. If they would buy with a small tweak, there are ways for Bad PMs to put up their credit card. Otherwise, ignore their advice, ignore mine, and keep pushing.

For the people who ask for more, yeah. There's always room for improvement. But if you take down people within our community, you can go sit and spin.

For a static personal site like this one, I’d recommend using caddy, which would’ve done most of these things out of the box.

nginx + ngx_pagespeed module also.

Talking about site speed comparisons, in order not to have results dependant on one's internet connection speed, would Firefox Dev Tool > "Responsive Mode" > throttling tool be a reliable way to benchmark sites ?

Anyone using this ?

(For example, test all sites on "regular 2g network" simulation and compare these times)

That's pretty much what it's designed for.

I think I remember reading about some esoteric differences with real life - related to dropped packets vs merely rate limiting the connection with HTTP2.

How would you actually strip unneeded javascript anyways? Let's imagine I have a setup like the OP and want to optimize it. Is there something, that would be able to strip down all JS functions, that are not needed by the site?

Webpack does dead code elimination through UglifyJS, see here https://webpack.js.org/guides/tree-shaking/

I've found link preload headers to be pretty effective at reducing the number of round-trips necessary to load my website.

Reply on news.ycombinator.com