Comments URL: https://news.ycombinator.com/item?id=48406640

Points: 296

# Comments: 120

It’s bad code smell and an anti-pattern as far as I’m concerned. Anyone who has actually has taken the time to learn POSIX/BASH shell to any degree or complexity knows that it’s fully capable of catching errors in-band and that a .sh file isn’t just a dumping ground for repeating your shell history.

> With a traditional CDN, if you set cache-control header, you can be sure that your files will be cached on the edge according to the header.

I’ve got a decade and a half of experience with Fastly, Akamai, and CloudFlare serving 100s of gigabits/sec of traffic. I can assure you this is not true at all. Cache-Hit ratio and Cache-Hit/Miss access time are highly situational and cache-control is best effort for every provider. No CDN will guarantee these values.

> With a traditional CDN, you can pick a hostname (Origin) and the file will be fetched from there. So, you can run a CDN directly over, say, a S3 Cloud bucket. Cloudflare works at your website domain level, and doing this is something like that is not possible.

“Cloudflare works at your website domain level.” This is poorly written, confusing, and fortunately, not at all true. You can CNAME i.example.com to an S3 bucket hostname and use i.example.com links on example.com. This also comes with some http/1.1 pipelining advantages and is a preferable way to architect.