Yeah, currently hosting LLHLS edge nodes in US + EU and caching CDN worldwide. The base cost grows if you have an audience of e.g. 2000 live viewers for a 2mbps stream = 4gbps.
Could be a lot cheaper and less need for global distribution if low latency weren't a requirement. And by low latency I mean the video stream you watch is ~2s behind reality just like Youtube, Twitch, Kick, etc. If your use case is VOD or can tolerate 10s latency streaming, single PoP works fine.
The point is that if I chose Vimeo or AWS/GCP/Azure for this, at their bandwidth pricing my (in my opinion modest) usage would be costing tens of thousands of dollars monthly, which it most certainly does not.
Managed service pricing and the perception of it needs to change, because it feels like a sham once you actually do anything yourself.
You misread the bandwidth cost part of my comment.
A $28/mo (Australian) vimeo subscription, or the "Advanced" $91/mo plan include the same 2TB bandwidth/month for viewers of your videos.
If you upload a 100MB video and it gets 20000 views the whole way through, you are now in the "contact sales" category of pricing.
This is why Youtube has a monopoly, because you've been badly tricked into thinking this pricing is fair and that 2TB is in any way shape or form adequate.
Written by somebody who hasn't taken 1 look at yt-dlp source code or issues. Google regularly pushes updates that "coincidentally" break downloaders. The obfuscation and things they do to e.g. break a download by introducing some breaking code or dynamic calculation required only part way through the video is not normal. They are not serving a bunch of video files or chunks, you need a "client" that handles all these things to download a video. At this point, if you assert that Google doesn't want to secretly stop it, you are either extremely naive, ignorant, or a Google employee.
Freedom with crypto means I can pay bills without unjust barriers, and no individual and especially no "financial institution" (ultimately the governments) decides if my transaction goes through or can confiscate the money without confiscating me.
With Paypal, Stripe, Visa, Mastercard, and all the payment gateways and controlled entities in between, you will be banned for a "trade secret" reason which they have no legal obligation to reveal. Example 1: https://kiwifarms.st/threads/payment-processor-censorship-vi... (You may discredit the source and ignore factual information and numerous examples at your own discretion)
When you lost the ability to pay for things, you can be starved, of food and more importantly your principles and dignity.
Like many other issues that I consider political, what is important to me and what I believe to be actually righteous in the end is more important than the issues of e.g. personal responsibility from being scammed, or criminal and money laundering transactions. Remember where the term "money laundering" came from.
Github is my push --mirror location, nothing more. Main is a popular Gitlab instance gitgud.io, and I host my own secondary mirror.
Gitlab is of course adding more AI and corpo garbage, and once they prevent disabling these "features" on community editions we'll see a fork of gitlab, probably.
The assertion that github is some bustling hub of opportunity is a strange one. At best you get people more likely to contribute because they already signed up, and a contribution from somebody not willing to sign up to another free service or simply email you an issue report is a contribution worth missing.
I always use a custom server (even express will work) with next.js, because I found the middleware and edge stuff a load of overcomplicated BS.
Client side works like regular react app, SSR for any pages where the initial props are just populated from the server site is easily controlled, and the whole system is simple to reason about.
There are other frameworks out there to do this now, but I'm comfortable with this and it just works, so no reason to change.
No. First of all, just check for `navigator.brave`. If it exists, it's Brave. When I ran a .onion site I added a JavaScript check and if navigator.brave was present, it redirected users to a specific page saying:
> Hey, there's something funny about your Tor Browser. When browsing Tor hidden services (.onion), you should be using Tor Browser. Are you using an outdated version, or perhaps something else entirely?
Brave is chrome. Tor browser is firefox, has a bunch of tweaks, different default settings, and a different fingerprint. Also when browsing on Tor, you should disable JavaScript as it's a source of many vulnerabilities.
Hey, funny to see my project mentioned here also. Yes, similar in concept.
Some differences:
- Uses HAProxy (duh)
- Proof of work can be either sha256 or argon2
- Optional recaptcha/hcaptcha in addition to the proof of work
- Includes a script for your page that will re-solve the challenge in the background before the cookie expires
There's also a control panel, dns server, etc. I kinda built my own everything because I refused to use bunny/cloudflare/whatever.
One thing I will say though, is that proof-of-work alone isn't a solution for ddos mitigation and bot protection! I've seen attackers using a mass of proxies and headless browsers to solve the challenge, or even writing code to extract and solve the challenge directly (https://github.com/lizthegrey/tor-fetcher). To adequately protect against more targeted attacks, you need additional acl and heuristics, browser fingerprinting, tls fingerprinting, ip reputation, etc. I do offer the whole thing setup as a commercial service, but will refrain from too much shilling.
It's fun, and I love seeing similar softwares help fight the horde of AI scrapers :^)
Offering more detailed timeouts for other stages of the request would be great, too.
For example with HAProxy you can configure separate timeouts for just about everything. The time a request is queued (if you exceed the max connections), the time for the connection to establish, the time for the request to be recived, inactivity timeout for the client or server, inactivity timeout for websocket connections... The list goes on: https://docs.haproxy.org/3.1/configuration.html#4-timeout%20...
Slowloris is more than just the header timeout. What if the headers are received and the request body is sent, or response consumed very slowly? And even if this is handled with a "safe" default, it must be configurable to cater to a wide range of applications.