[embedding-shape]

I don't know exactly what the website was, but if it's just HTML, CSS, some JS and some images, why would you ever host that on a "pay per visit/bandwidth" platform like AWS? Not only is AWS traffic extra expensive compared to pretty much any alternative, paying for bandwidth in that manner never made much sense to me. Even shared hosting like we did early 00s would have been a better solution for hosting a typical website than using AWS.

[throwup238]

Especially when Cloudflare Pages is free with unlimited bandwidth, if you don't need any other backend. The only limit is 100 custom domains and 500 builds per month in their CI/CD, the latter of which you can bypass by just building everything in Github Actions and pushing it to Pages.

[fragmede]

Because Cloudflare pages has this doofy deploy limit hanging over your head. Even if you won't reasonably run into it, it's still weird. R2's free limits are significantly higher.

[anamexis]

AWS CloudFront pricing seems pretty competitive with other CDNs, at least for sites that are not very high traffic.

https://aws.amazon.com/cloudfront/pricing/

[embedding-shape]

> AWS CloudFront pricing seems pretty competitive with other CDNs

Sure, but it's unlikely you actually have to place a CDN in front of your manual, it's mostly text with few images. People default to using CDNs way too quickly today.

[jazzyjackson]

But why host my own static content when cloudflare r2 buckets will give me a million reads a month for free?

[anamexis]

The whole point of CDNs is to host static assets, why wouldn't you use one? They are dead simple to use.

[embedding-shape]

Because caddy/nginx/apache (any web server really) can serve that content as well as any other? Better question is; why default to using more things before you actually need them?

Personally, software engineering for me is mostly about trying to avoid accidental complexity. People obsessing about "web scale" and "distributed architecture" before they even figured out if people actually want to use the platform/product/tool they've used tends to add a lot of complexity.

[everforward]

> Because caddy/nginx/apache (any web server really) can serve that content as well as any other?

That's not really true if you care about reliability. You need 2 nodes in case one goes down/gets rebooted/etc, and then you need a way to direct traffic away from bad nodes (via DNS or a load balancer or etc).

You'll end up building half of a crappy CDN to try to make that work, and it's way more complicated than chucking CloudFlare in front of static assets.

I would be with you if this was something complicated to cache where you're server-side templating responses and can't just globally cache things, but for static HTML/CSS/JS/images it's basically 0 configuration.

[mystifyingpoi]

> That's not really true if you care about reliability

While reliability is always some concern, we are talking about a website containing docs for a nerdy tool used by a minuscule percentage of developers. No one will complain if it goes down for 1h daily.

[anamexis]

How is setting up a web server not using more things than you need when you could just drag and drop a folder using one of many different CDN providers? (Or of course set up integrations as you want)

[embedding-shape]

Just because you're using a UI doesn't mean it isn't more complicated. I'm thinking "complexity" in terms of things intertwined, rather than "difficult for someone used to use GUIs".

[DiabloD3]

Far more expensive than just having a dumb server somewhere at some normal host.

People simply do not understand how expen$ive AWS is, and how little value it actually has for most people.

[nijave]

It's really a tradeoff of saving time by paying more money. A lot of people chose it when they'd rather not pay more money and end up unhappy

A lot of other people also pick it for very narrow use cases where it wouldn't have been that much more time to learn and do it themselves and end up paying a lot of money and also aren't happy

It's pretty nice for mid-size startups to completely ignore performance and capacity planning and be able to quickly churn out features while accumulating tech debt and hoping they make it long enough to pay the tech debt back

[dandeto]

A year ago I researched this topic for a static website of my own. All providers I looked at were $5 and I want to say the cheapest I found was slightly lower. By comparison, I am still within free tier limits of AWS S3 and cloudfront (CDN) since I am not getting much traffic. So my website is on edge locations all over the world as part of their CDN for free, but if I host on a single server in Ohio it costs $5/month.

[massysett]

Did you check NearlyFreeSpeech.net? If you are getting little traffic it costs practically nothing for a static site.

[anamexis]

Most people will never make it past the free tier on any of CloudFront, Cloudflare, Netlify, Render, etc.

You can just drag and drop a folder and have a static site hosted in a few minutes.

[SatvikBeri]

Seems like the maintainer just tried AWS out of curiosity, and never needed to optimize hosting until scrapers suddenly slammed the site.

[tarsius]

That's exactly what I did! Though I would call it morbid curiosity. :P

After initial setup it was smooth sailing. Other more reasonable setups would also have been smooth sailing, but... they weren't setup yet. I was uneasy about the possibility of a surprise bill happening, as it eventually did, but until the brain dead LLM leeches came along, that just never happened. After a decade of it not happening, I wasn't that concerned anymore, but I guess when it comes to the AI bots, I had my head in the sand a bit. I still though something like a 500% bill might happen, not 5000%.

Once it did happen, I immediately shut my sides down, and within the hour the account was no more. On the way out I saw that you can now actually set a "spending limit", it still had a [new] next to it. I tried setting it up, but could only quickly figure out how to setup a notification. It might be possible to set an actual spending limit, but not in a few minutes -- probably got to read some documentation for that.

But even if this were a one click setting, it wouldn't have made a difference at this point. You do this once and I am gone. Also, I wanted to move away from Amazon anyway, so really, this was the kick in the pants that I needed.

For now I am using Github Pages for the very static parts, and the free hosting provided by my email provider, for the slightly less static manuals generated with Github Actions. I would have made sense to use Github for both (not least so that Microsoft could cover the cost of the bots they have unleashed), but I wanted to avoid the complexity of committing to the same pages repository from the CI pipelines of multiple package repositories.