Matthias247

11 ms·

91.
by Matthias247 4y ago
HTTP/3 uses QUIC, which mandates TLS. This, and the general protocol specification which requires peers to store very detailed state about every packet sent, make it much more resource intensive than HTTP/1.1 over TCP. Source: I i
92.
by Matthias247 4y ago
Maybe in the open source world? In some closed source company codebases that are known to me, there were still million lines of code on JDK8, since efforts to port everything to a newer JDK don't necessarily enjoy a high priority. As a
93.
by Matthias247 4y ago
But they require a new JDK version, and a surprisingly high amount of Java projects are still stuck on JDK8.
94.
by Matthias247 4y ago
Nothing. It would be a property of the webserver whether multiple requests can be processed in parallel. And a HTTP/3 server which can't would most likely be broken. It it would not process other QUIC packets while handling one re
95.
by Matthias247 4y ago
it's actually the perfect resolution for this size - since it finally gets you 2x scaling with "previously native" font sizes. we started with around 100dpi when 1920x1200 screens where trendy. Font and UI sizes on those were
96.
by Matthias247 4y ago
if you want 220dpi, then any TV isn't an option. Even if you just want 160dpi, then current 27" 4k screens are your only option.
97.
by Matthias247 4y ago
I don’t think this matters at all. People usually are not writing their own http libraries. They use existing ones, and those already take care of reading the http headers and the body separately. One should avoid handwriting a http server
98.
by Matthias247 4y ago
The vscode codebase might be interesting.
99.
by Matthias247 4y ago
perf can do the job on Linux. Networking cost will be within callstacks for the system calls that send and receive packets (sendmsg, sendmmsg, recvmsg, recvmmsg). Crypto cost in functions which sound like the crypto primitives being used. T
100.
by Matthias247 4y ago
My experience as former technical lead of a major HTTP/3 deployment is that while TLS certainly adds CPU overhead, it's by far not as major as the overhead of TLS on top of TCP is. QUIC in general is far less efficient than TCP+TL
101.
by Matthias247 4y ago
File IO isn’t epoll-able. The operations are always blocking. Those could however be offloaded onto a threadpool, to avoid the blocking to affect any other requests that are processed by the same Nginx worker. Nginx however only partially d
102.
by Matthias247 4y ago
> The documentation is kind of vague, but apparently you have to re-enable it regularly.[3] This is correct. And in the end it means more or less that setting the socket option is more of a way of sending an explicit ACK from userspace t
103.
by Matthias247 4y ago
Enabling it will lead to more ACK packets being sent, which leads to lower efficiency of TCP (the stack spends time in processing ACK packets) and lower link utilization (these packets also need space somewhere). My thought is that the beha
104.
by Matthias247 4y ago
TCP is always reliable, the choice of this algorithm will never impact this - it will only impact performance (bandwidth/latency) and efficiency. Enabling nagle by default will lead to elevated latencies with some protocols that don&#x
105.
by Matthias247 4y ago
The article is way too opinionated about „golang is doing it wrong“ for a decision that neither has a right or wrong. Nagle can make sense for some applications, but also has drawbacks for others - as countless articles about the interactio
106.
by Matthias247 4y ago
Dividing packets into MTUs is the job of the tcp stack - or even the driver or NIC in the case of offloads. Userspace software shouldn’t deal with MTUs and always use buffer sizes that make sense for the application - eg 64kB or even more.
107.
by Matthias247 4y ago
You can configure whether the cache key includes a particular header or query parameter in a lot of CDNs. So as long as your user identify is transmitted in one of those, it would work.
108.
by Matthias247 4y ago
have you really bought it, in terms of paying extra money besides the prime costs just for the show? Or just subscribed to prime? Since you put "bought" in quotes I guess it could mean both. And I think LeoPanthera was referring t
109.
by Matthias247 4y ago
I'm using a pretty low level Displayport switch, and I think it also takes 4s to get a picture back. I guess the reason for this is actually the monitor and GPU. They will resync to each other after figuring out the peer capabilities a
110.
by Matthias247 4y ago
I've worked as a principal engineer at Amazon before, and I can tell that testing and quality there isn't looked down upon. You can definitely get promoted just for improving products and never building any new feature. However it
111.
by Matthias247 4y ago
I think it's rather the opposite: If it's a very generic position, to which a lot of applicants could match. In this case big companies often do hiring events, where a lot of applicants are interviewed at the same day, and potenti
112.
by Matthias247 4y ago
Sure! Obviously you can only coalesce packets which have the have the same destination
113.
by Matthias247 4y ago
> though Go only preempts at function boundaries IIRC I think this was improved with more recent Go versions and now it is really preemptive
114.
by Matthias247 4y ago
I neither have experience with wireguard nor x/crypto. But I did some measurements on crypto algorithms for QUIC efficiency before. There I achieved about 350MB/s throughput on a single Ryzen 3900X CPU core using ChaCha20, compare
115.
by Matthias247 4y ago
It will likely not help a lot, because syscall overhead is not the bottleneck - but the implementation of the actual syscalls is - which is very visible in the flamegraphs. Both sendmmsg/recvmmsg and io_uring only help with the syscall
116.
by Matthias247 4y ago
It can not only be a lot faster, it definitely is. I did a lot of work on QUIC protocol efficiency improvements in the last 3 years. The use of sendmmsg/recvmmsg yields maybe a 10% efficiency improvement - because it only helps with re
117.
by Matthias247 4y ago
Answering it here would take some of the surprise away. But I will say it's totally worthwhile to get one of the puzzles if one enjoys puzzles in general. The mechanism of the ending seems to be the same in all the 6 puzzles, so you ca
118.
by Matthias247 4y ago
data corruption is slightly orthogonal to data loss. If your filesystem silently corrupts some data in a way you don't notice, you might just replicate the corrupted data to your backup system over time. And in that case you will still
119.
by Matthias247 4y ago
I would not. I don't want to be concerned with whether any of this is fake money or how to detect it and hugely prefer an electronic transfer.
120.
by Matthias247 4y ago
bank transfers are usually free in europe

More ›