Cloudflare is experiencing elevated 5xx responses
appears to be down for many websites see https://www.cloudflarestatus.com/
42 comments
Apropos nothing, I found out today that bash.org died. https://news.ycombinator.com/item?id=37295238 was the last I was able to find about it, and it's the top google search result.
I've been thinking of the best way to make a website that won't die when I die. It seems like a hard problem. I was originally going to just trust that Github Pages will be around forever, but I'm not so sure anymore.
Running your own server is fraught with danger. In theory it's just a matter of making sure your account has enough credits to last 100 years. In practice, I can't count the number of times that servers have had one problem or another that requires manual intervention to resurrect.
It's surprising that this isn't a solved problem. "Put some HTML somewhere" has always left the question of "where?"
Maybe paying for your own S3 account and shoving the HTML onto that is the way to go. Still, thinking of the number of companies that have survived even 20 years, the odds aren't good that both S3 and Cloudflare will run flawlessly for decades.
Oh yeah, the reason we're able to read these comments now is because Dan migrated HN away from Cloudflare, I think to Route 53. So maybe S3 + Route 53 has the highest odds of standing the test of time.
Then there's the question of what to write. But at least "test website please ignore" has a chance of lasting a century, unlike everyone else.
I've been thinking of the best way to make a website that won't die when I die. It seems like a hard problem. I was originally going to just trust that Github Pages will be around forever, but I'm not so sure anymore.
Running your own server is fraught with danger. In theory it's just a matter of making sure your account has enough credits to last 100 years. In practice, I can't count the number of times that servers have had one problem or another that requires manual intervention to resurrect.
It's surprising that this isn't a solved problem. "Put some HTML somewhere" has always left the question of "where?"
Maybe paying for your own S3 account and shoving the HTML onto that is the way to go. Still, thinking of the number of companies that have survived even 20 years, the odds aren't good that both S3 and Cloudflare will run flawlessly for decades.
Oh yeah, the reason we're able to read these comments now is because Dan migrated HN away from Cloudflare, I think to Route 53. So maybe S3 + Route 53 has the highest odds of standing the test of time.
Then there's the question of what to write. But at least "test website please ignore" has a chance of lasting a century, unlike everyone else.
Long ago, Michael O'Connor Clarke wrote "The KillSave Manifesto" for me. I hosted it for a while via 1and1, and eventually we won, and save wasn't an explicit required step in most cases any more.
Sadly, Michael died long before his time, and I'm not going to be around for ever, but I suspect you'll be able to read his work for a very long time via the Internet Archive at
Sadly, Michael died long before his time, and I'm not going to be around for ever, but I suspect you'll be able to read his work for a very long time via the Internet Archive at
https://web.archive.org/web/20080621173441/http://killsave.org/Killsave_Manifesto.htm
And that you throw a thousand or two dollars their way in your will. Pay it forward.
> I've been thinking of the best way to make a website that won't die when I die
It's said that most people look at a painting in a gallery for fewer than 30 seconds. I'd be willing to guess that people are going to look at an old HTML document even less than that (not to mention web scrapers which will be the most likely to uncover an old page)
There is no guarantee that the web services, companies, technologies, or even the Internet itself will stick around forever. Everything dies. The point, however, isn't to try to leave a permanent mark somewhere that says "I was here". Instead, one should nurture a sense of stewardship for newer generations to help instill them with good values that they will carry forwards in turn
It's said that most people look at a painting in a gallery for fewer than 30 seconds. I'd be willing to guess that people are going to look at an old HTML document even less than that (not to mention web scrapers which will be the most likely to uncover an old page)
There is no guarantee that the web services, companies, technologies, or even the Internet itself will stick around forever. Everything dies. The point, however, isn't to try to leave a permanent mark somewhere that says "I was here". Instead, one should nurture a sense of stewardship for newer generations to help instill them with good values that they will carry forwards in turn
> It's said that most people look at a painting in a gallery for fewer than 30 seconds.
So I guess the lesson is, if you want something to stick around, go for good old fashioned paper. It doesn’t need electricity, it doesn’t need servers, software, maintainers, infrastructure or any of that. And if taken care of, it might last a thousand years give or take.
So I guess the lesson is, if you want something to stick around, go for good old fashioned paper. It doesn’t need electricity, it doesn’t need servers, software, maintainers, infrastructure or any of that. And if taken care of, it might last a thousand years give or take.
Alternatively, you could just stick it into a 3,000 lb sarcophagus, like what this guy did for a bag of Flaming Hot Cheetos
https://www.youtube.com/watch?v=d6_C7z4XvOQ
https://www.youtube.com/watch?v=d6_C7z4XvOQ
Humanity have traditionally solved this problem by having children and putting in their will "please make sure my HTML page is still up and running until 2123".
There’s always the expensive option:
Open a trust with a major bank to preserve that website and fund it with a long term annuity or investment, with instructions on how to execute it.
Not perfect of course, as trust firms and banks also fail — but I think that’s the best chance to reach 50-100 years after death.
For $200 per year, with $100 fees and $100 hosting costs, you’re looking at a $5,000 investment — assuming the “safe” 4% yearly drawdown.
Open a trust with a major bank to preserve that website and fund it with a long term annuity or investment, with instructions on how to execute it.
Not perfect of course, as trust firms and banks also fail — but I think that’s the best chance to reach 50-100 years after death.
For $200 per year, with $100 fees and $100 hosting costs, you’re looking at a $5,000 investment — assuming the “safe” 4% yearly drawdown.
Wordpress has a plan just for that. https://wordpress.com/100-year/
For a one time payment of: US$38,000.
Even if your page lasted 100 years, your cert wouldn’t. Who knows what DNS, or the equivalent to IP looks like in the 22nd century.
http://paulgraham.com actually fails if you try to access it via HTTPS. Maybe that's one way to get around the problem.
Is there really no way to date a cert for a century?
Is there really no way to date a cert for a century?
Nope, they top out at 398 days. I think there's even been some talk about limiting them to 90 days.
Is there an actual RFC for that? I was thinking it was just browsers (Google really) who enforced it by convention.
As I understand it, it is just browsers enforcing it, with Safari [1] doing so first, then Chromium [2] doing the same, both doing so in 2020.
[1] https://support.apple.com/en-us/102028
[2] https://chromium.googlesource.com/chromium/src/+/master/net/...
[1] https://support.apple.com/en-us/102028
[2] https://chromium.googlesource.com/chromium/src/+/master/net/...
Cloud providers have automated cert renewal.
Hmm. This is interesting. Skimming over https://docs.aws.amazon.com/acm/latest/userguide/managed-ren... it sounds like it might be a matter of using DNS validation, whatever that is.
So that just leaves the question of registrar. As much as I like gandi, it seems like a better idea to migrate the domain itself to Amazon too. https://docs.aws.amazon.com/Route53/latest/DeveloperGuide/re...
Theoretically, this would be a self-contained system — the only point of failure is AWS itself. As long as you load your account with a few thousand bucks, perhaps it will last “forever”.
I also like pg’s philosophy of "just pretend https doesn’t exist", but if it’s guaranteed that the automation won’t fail, it seems gratuitously fussy to insist on plain http.
Thanks!
So that just leaves the question of registrar. As much as I like gandi, it seems like a better idea to migrate the domain itself to Amazon too. https://docs.aws.amazon.com/Route53/latest/DeveloperGuide/re...
Theoretically, this would be a self-contained system — the only point of failure is AWS itself. As long as you load your account with a few thousand bucks, perhaps it will last “forever”.
I also like pg’s philosophy of "just pretend https doesn’t exist", but if it’s guaranteed that the automation won’t fail, it seems gratuitously fussy to insist on plain http.
Thanks!
A. Write it down sufficiently succinctly to publish in academic journals. Or failing that, a book with an ISBN that gets archived in libraries. Perhaps creative commons etc licencing will help a bit, but it's gotta make it to distributed archives.
B. Teach a future generation. It lives through them and gets built on.
C. Embed it in what you do. It might not be a webpage, but there's code and ideas living in organisations today that have outlived their authors.
In this sense, perhaps bash.org hasn't died. It's just outlived its domain.
B. Teach a future generation. It lives through them and gets built on.
C. Embed it in what you do. It might not be a webpage, but there's code and ideas living in organisations today that have outlived their authors.
In this sense, perhaps bash.org hasn't died. It's just outlived its domain.
Maybe the Long Now Foundation could start a 10k years hosting plan. That would be very interesting. Also being a sysadmin for a 10k server park would be a interesting job.
> Also being a sysadmin for a 10k server park would be an interesting job.
Ideally not frequently interesting, though.
Ideally not frequently interesting, though.
journalctl —follow -u long-now.service
https://posthaven.com/
Says it supports full HTML theming so you could have ~arbitrary content.
Says it supports full HTML theming so you could have ~arbitrary content.
posthaven seems like a good shot: https://posthaven.com
See faq>archive mode
Of course it’s nothing more than an organization of people promising you that they will keep the lights on for you insofar as it is feasible, but i’m not sure you can get better than that in practice
Of course it’s nothing more than an organization of people promising you that they will keep the lights on for you insofar as it is feasible, but i’m not sure you can get better than that in practice
> I've been thinking of the best way to make a website that won't die when I die.
I've spent a lot of time thinking about the same, and eventually had to ask myself: why do I care? (This is just my personal reflection, not an indictment of your goals. I'm a fatalist.)
If I have to go to the trouble of setting up a 100-year trust to maintain a digital presence after my death because nobody else finds what I have to say compelling enough to archive on its own merit, is it worth preserving? Or am I just wasting my time, adding more debris to float around in [cyber]space?
Blockchain would have been useful here. Everyone could share in the storage of collective, immutable history. But all of it is lost when some trillionaire inevitably decides to start a competing, incompatible product with a flashier UI.
I've spent a lot of time thinking about the same, and eventually had to ask myself: why do I care? (This is just my personal reflection, not an indictment of your goals. I'm a fatalist.)
If I have to go to the trouble of setting up a 100-year trust to maintain a digital presence after my death because nobody else finds what I have to say compelling enough to archive on its own merit, is it worth preserving? Or am I just wasting my time, adding more debris to float around in [cyber]space?
Blockchain would have been useful here. Everyone could share in the storage of collective, immutable history. But all of it is lost when some trillionaire inevitably decides to start a competing, incompatible product with a flashier UI.
For fun, of course. The heat death of the universe means it’s the only true motive. Holding an uptime world record is pretty compelling, at least to me.
Make it profitable, or at least sustaible.
Eh, websites (and all things) come and go.
As long as someone cares to maintain them, they will be there. And if nobody cares to maintain them, well, may others rise in their place.
https://en.wikipedia.org/wiki/Impermanence
As long as someone cares to maintain them, they will be there. And if nobody cares to maintain them, well, may others rise in their place.
https://en.wikipedia.org/wiki/Impermanence
IPFS perhaps?
I keep imagining an elevated 5xx response is a 599 when it should be a 503 or a 504.
If status codes bleed into 6xx we'll know HE COMES.
I am starting an early warning system, a daemon process, polling random web sites for status lines like this. Do I need to handle Unicode?
I am starting an early warning system, a daemon process, polling random web sites for status lines like this. Do I need to handle Unicode?
Raised to the power.
Discord is having issues right now, unsure if it's related.
- Investigating - We are working to correct an issue that is causing some servers to be degraded and unavailable. - Update - We are continuing to investigate this issue.
- Investigating - We are working to correct an issue that is causing some servers to be degraded and unavailable. - Update - We are continuing to investigate this issue.
It's fixed, at least for us.
We had a bunch of 530 codes returned on our endpoints, although Cloudflare says that they implemented a fix at 22:39 UTC, we don't have an increased rate of errors since 22:10 UTC.
We had a bunch of 530 codes returned on our endpoints, although Cloudflare says that they implemented a fix at 22:39 UTC, we don't have an increased rate of errors since 22:10 UTC.
Makes sense, they waited until the fix was fully deployed and working before saying that it had been implemented.
Are databases of Cloudflare's core services deployed in more than one AZ yet?
Ah, that might explain why I was seeing http 520 errors from various sites. It was the first time in a long time that I needed to lookup a http response code, only to find it's an unofficial one.
Didn’t experience any outage for mailsnag.com. We have newrelic synthetic checks around the globe, and another uptime monitor. Not even a blip on the charts.
I'm also seeing a lot of 1016/1018 "Origin DNS errors" across a variety of web services. I saw this on my own company's website first around 21:40 UTC, although it's not bad enough to materially affect our traffic volume.
Tell HN: