And you almost certainly don't need one for scaling purposes. I've handled frontpage traffic multiple times without.
What I wish existed (or maybe does) is something like a CDN with ~8 global locations that promises to keep your static site in a warm cache for a small fee, without me having to set up a fly.io app manually. That would let my blog always load fast anywhere on the planet even for the first hit in a region.
[1] https://support.bunny.net/hc/en-us/articles/360017093479-Und...
cloudflare cache reserve or bunny.net as someone else mentioned
IIRC this is what Cloudflare Argo is supposed to do? https://www.cloudflare.com/en-gb/application-services/produc...
Rather than routing cache misses to origin, it'll route to other PoPs with the cached content?
Depend on external services, die by external services. If you can’t afford redundancy, then you must accept downtime.
I build all of my dependent libraries locally and bundle with my app. For one application, I maintain and build the entire OS from source. I depend on aws,azure cloud storage for my own hosted cdn. I can fail back to my own hosts if necessary.
I would have thought the risks from 3rd party production dependencies would make devs do all in their power to minimize them. But IIUC it's a somewhat common practice.
Is there something about the market dynamics that make this a reasonable tradeoff? E.g., the problems are rare and this somehow allows faster time to market?
(Asking as a systems programmer, very much not a web developer.)
I don’t have any data but I really don’t think it actually is that common. I definitely heard it used a lot more in the 2000s jQuery era but not really today. Browsers even have isolated storage these days so the benefits really are infinitesimal.
Tell that to any developer (not just web) that outsources everything these days e.g. authentication, storage, analytics, database, etc - you name it.
They think it's less risk (and this includes managers and directors in larger companies). Apparently in-house would take forever, cost more and be unable to maintain anything these days.
I don't get it either but it's the current train of thought / trend and is not tied to individual developers.
Yep. Entire teams work on these products. You have a couple of options:
1. Use them
2. Make it internally with less staffing, resulting in an inferior product
Option 2 is viable if you really only need a small subset of features/deep customization, but if you run into the edges the problem space will explode on you. At that point, you either have to throw more resources at it or go 3rd party.
I also miss being able to host random CSS and assets on Tumblr.
Otherwise, it should be on my own CDN. Especially for enterprise sites.
There is also a tool to simplify migration https://www.jsdelivr.com/unpkg
Browsers have partitioned caches per origin for over 10 years, so there's no performance gains from getting cache hits from other websites.
in the strictest sense, unpkg is not a cdn. unpkg is a javascript library hosting service that uses cloudflare as a cdn. cloudflare is not down, therefore this headline "unpkg cdn down..." reads as a massive "oh shit!" to infra/devops type cloudflare customers who then have to immediately worry that their site is down, only to realize no, they're fine, this is just devs using the term to mean a different thing (roughly: "third party javascript host").
Do they? Same to me.
> in the strictest sense, unpkg is not a cdn. unpkg is a javascript library hosting service that uses cloudflare as a cdn.
This is just arguing over semantics. It even says on the website that unpkg CDN is powered by Cloudflare.
By your analogy if a website hosted on AWS is down, is the whole of AWS down? If I provide managed kubernetes on GCP and my service is down, does that make GKE or GCP down? No.
A CDN is a CDN just like a website is a website, whether it's backed by Wordpress, NextJs or any other thing.
Unpkg runs on Fly with Cloudflare as CDN.