3 ms·
True, it's probably not fair to extrapolate these results to every VPN installation, but I'd argue that retail VPN providers are among the most widely deployed
by vpnintel 6y ago
True, it's probably not fair to extrapolate these results to every VPN installation, but I'd argue that retail VPN providers are among the most widely deployed WireGuard and OpenVPN installations (are there bigger ones I'm not thinking of?), and that they are probably pretty optimized for the advantage it'd give to be faster and/or operate at lower cost than their competitors.
- foolmeonce 6y ago> they are probably pretty optimized for the advantage it'd give to be faster and/or operate at lower cost than their competitors. I can't agree with that in this context. People have little idea what they are doing and what they should choose and follow various guides of various ages, pushing people who select an older implementation to select newer algorithms than the past defaults risks negative publicity. Selling rather slow service pretty cheap is how you get the volume in virtually any service market, most users making a choice preventing them from using the network bandwidth they were sold is also a bonus for many companies with inadequate peering. If the VPN was selecting for performance, why wouldn't they select the null algorithm? In wireguard you seem to have no choices, but in normal IPsec you could make all sorts of different choices that have different tradeoffs, but in general: the weaker and more open to attack, the faster and cheaper your offering can be and the more likely one blog entry by the wrong person kills your company.