I find it mildly embarrassing that people still manage to screw up PMTUD, a big CDN, no less ...
I find it mildly embarrassing that people still manage to screw up PMTUD, a big CDN, no less ... but then again, I guess I appreciate that they did actually care to fix it once they noticed, which is not something to take for granted, and at least it was not the idiotic idea of filtering ICMP (has Amazon ever managed to fix EC2?), but an actual oversight. However, that blog article seems somewhat misinformed: In addition to tunnels, one major reason for PMTUs smaller than the standard Ethernet MTU of 1500 are DSL connections based on PPPoE, both for IPv4 and for IPv6. The only reason most people don't notice the breakage caused by incompetent admins (in the let's filter ICMP case) is because home routers tend to mess with the TCP options in SYN packets, adding an MSS matching the MTU of the DSL connection. Of course, that's a braindead fix, as it doesn't work for anything other than TCP. BTW: I doubt it was that small a number of users that noticed, a customer of mine also noticed (native IPv6 over ADSL), but given that this usually is caused by intentionally clueless admins, I just told them that the server admin probably was incompetent rather than reporting the problem to cloudflare ...
Posted by zAy0LfpBZLC8mAC on Feb 5, 2015, 12:47 AM
Tags
Related issues
- I mean that's the problem though.
- I've always believed the opposite: get conditionals deep in your code so that the higher level control flow is regular.
- The biggest problem with AI isn't being able to do things.
- I am a 3D printing maker type nerd (I make my own tools for other creative tasks, rather than as their own outcomes) and I saw an incredi...
- Of course not, those are clearly scams.
- Claude makes more detailed plans that seem better if you just skim them, but when analyzed, has a lot of errors, usually.
Raise real-world problems, find people facing the same issue, and discuss solutions.8,000 issues0 comments
hackernews (EdTech)
I mean that's the problem though.
I've always believed the opposite: get conditionals deep in your code so that the higher level control flow is regular.
The biggest problem with AI isn't being able to do things.
I am a 3D printing maker type nerd (I make my own tools for other creative tasks, rather than as their own outcomes) and I saw an incredi...
Of course not, those are clearly scams.
Claude makes more detailed plans that seem better if you just skim them, but when analyzed, has a lot of errors, usually.
Effectively everyone is building the same tools with zero quantitative benchmarks or evidence behind the why / ideas … this entire s...
> creating and operating a small company in Germany is a nightmare To be fair, I think the problem operating in Germany is its federated...
I find it mildly embarrassing that people still manage to screw up PMTUD, a big CDN, no less ...
I find it mildly embarrassing that people still manage to screw up PMTUD, a big CDN, no less ... but then again, I guess I appreciate that they did actually care to fix it once they noticed, which is not something to take for granted, and at least it was not the idiotic idea of filtering ICMP (has Amazon ever managed to fix EC2?), but an actual oversight. However, that blog article seems somewhat misinformed: In addition to tunnels, one major reason for PMTUs smaller than the standard Ethernet MTU of 1500 are DSL connections based on PPPoE, both for IPv4 and for IPv6. The only reason most people don't notice the breakage caused by incompetent admins (in the let's filter ICMP case) is because home routers tend to mess with the TCP options in SYN packets, adding an MSS matching the MTU of the DSL connection. Of course, that's a braindead fix, as it doesn't work for anything other than TCP. BTW: I doubt it was that small a number of users that noticed, a customer of mine also noticed (native IPv6 over ADSL), but given that this usually is caused by intentionally clueless admins, I just told them that the server admin probably was incompetent rather than reporting the problem to cloudflare ...
Comments (0)
No comments yet. Start the discussion.