Cloudflare is a regular problem for Starlink users. We're on CGNAT so users share IPv4 addresses. I see CAPTCHAs when using Starlink ten times as often as on my other ISP. I don't think it actually breaks things the way this article describes, it seems like a gentler behavior, but it's annoying.
A few months ago I got on Akamai's naughty list (with my other ISP) for some very light automated website downloading. That was a straight block with HTTP errors and I had to use a proxy to access the Web. It cleared up after a few days.
The lack of any user feedback or support for this situation is really annoying. Reminds you how much power the CDNs have. It'd be really bad if loading websites got as difficult as sending email through all the layers of spam filtering.
I feel like Starlink could at least partially mitigate this by supporting IPv6. T-mobile US supports IPv6, and I hardly notice this as an issue on my phone. Or the time my work ran the business over a 4G mobile while waiting for ISP install.
A genuine question from an ignoramus: how on earth did Starlink launch a brand new ISP in 2020 which doesn't support IPv6? Is IPv6 really so difficult? Does actually nobody care about IPv6 still, after all these years?
Not an answer to your question, but an indicator of shared culture: Tesla vehicles also don’t support IPv6 whatsoever.
Things you might use an internet connection for in your Tesla include triggering air con remotely, live traffic and satellite maps, streaming music or online radio, web browsing, or YouTube/Netflix/Disney+ clients.
It completely refuses to use IPv6 over mobile or wi-fi. Also it refuses to access anything over IPv4 (apart from DNS) which resolves to an RFC1918 address, even if it's connected to said RFC1918 network.
So yes, Starlink and Tesla are different companies, but I see cultural parallels which I'm sure surprises nobody.
I've been wondering that too, it's kind of confounding. Particularly since Starlink started with CGNAT to manage a limited IPv4 address allocation.
FWIW there've been hints from time to time that Starlink was working on IPv6; users reported being given working addresses. That mostly stopped though when they handed over ISP operations to Google last year.
> Cloudflare is a regular problem for Starlink users. We're on CGNAT so users share IPv4 addresses. I see CAPTCHAs when using Starlink ten times as often as on my other ISP. I don't think it actually breaks things the way this article describes, it seems like a gentler behavior, but it's annoying.
I've been noticing this too, and it's why Starlink remains my secondary ISP/bulk transfer connection. If I had to drop one connection, I'd drop Starlink for this reason alone.
There are some sites that I simply can't browse, and it's not Cloudflare errors, either. Lowes, in particular, simply returns error pages for anything but the main landing page on a regular enough basis. Of course, my observed public IP changes so it's not consistent, but it's genuinely annoying.
> I've been noticing this too, and it's why Starlink remains my secondary ISP/bulk transfer connection. If I had to drop one connection, I'd drop Starlink for this reason alone.
Could cloudflare legally charge them a bribe to captcha their users less? It isnt good to have a company in this position of power if so.
Because my other connection is a 25/3 WISP link that mostly doesn't. I generally see about 5/1 in the evenings, if that.
I've had several area WISP connections, as there's no wired infrastructure to my area, and they vary in quality. I work full time remote, so I need two connections as a general habit - I can work with one, but when that one is down for a week straight, I have problems. I like being able to fail over.
I typically keep one connection for "interactive" traffic, and one for "bulk transfer/failover" - things like my local Ubuntu repo mirror, offsite backup traffic, etc. And I can fail to it if needed, which I do often enough.
On a good day, Starlink is far better than my WISP connection, and I have some machines routed out it persistently. On a bad day, I can't hit much from it, because that particular public IP has been blocked from large parts of the internet. It's very hit and miss, and overall bandwidth has definitely dropped from the early days, though reliability of getting packets where they need to go is drastically improved.
Starlink supports IPv6 addresses, I'm sure it would help. My network infrastructure is just lacking general IPv6 support, as I've not cared to get it set up in great detail, and my testing has demonstrated that my IPv6 addresses behind my router are publically reachable, so... I'll get around to proper firewalling at some point.
I could do a range of things to solve it, but as I have two ISPs, I typically just switch to the one that works better for the solution. I'm aware it's not a technically fancy solution, but it's quick and easy (change the gateway on the machine), and works fine.
A combination of shotscraper and metascraper; really more web previews than archives. And in a single thread, to different hostnames, maybe one every 10 seconds? Honestly surprised Akamai or anything even noticed. I fake my user agent now, lesson learned.
But any automated tool won't work. I have a similar problem with my self hosted feed reader, my vps hosting ip doesn't have 100% reputation with Cloudflare and I can't download some feeds
I'm really hoping cloudflare gets busted for having backroom deals with big ISPs or something. (For instance, if the cgnat had a cloudflare CDN cache endpoint behind / accociated with it, I suspect the IP would be white listed.)
They're working on double dipping by providing both the problem and the solution. Somehow this is not a recurrent issue for every other CDN / ddos shield. They're not even mentioning any other hosting company collaborating on this open solution that requires hardware from a specific company they totally don't have a deal with...
A few months ago I got on Akamai's naughty list (with my other ISP) for some very light automated website downloading. That was a straight block with HTTP errors and I had to use a proxy to access the Web. It cleared up after a few days.
The lack of any user feedback or support for this situation is really annoying. Reminds you how much power the CDNs have. It'd be really bad if loading websites got as difficult as sending email through all the layers of spam filtering.