Coup de Grace

君技术本当上手 2004

标题算 neta 一下日本语本当上手.

最近智齿发炎了两次,实在挺不住了,决定消炎后去拔牙.

停更的几个月实在是疏于阅读

对一些糟心的人和事做不到移开目光的话,暴力破解+永远不要降低自己的水准.

都会使用 Linnk 将文章快照免得丢失.


我的编辑器选择

我实际已经从 MarkEditor 迁移到 VSCode 又迁回 MacDown很久了.

主要是因为实在不喜欢在我的工程里出现不明不白的.config文件夹.

最近看到 me 在开始推广,想必他们的课卖的不太好吧.

VSCode 自动 hot-reload 后老是出现问题

有时候居然还需要 reinstall…

是质量不如以前了还是其他原因就不得而知了.


Comparing HTTP/3 vs. HTTP/2 Performance

地址

这就很可玩味了.

基于 UDP 去实现,首先对比了time to first byte (TTFB).

拥堵算法则是Reno/CUBIC.

性能方面的其他优点就没有了..

下方评论里一位头像敏感的也指出:

Since this is a UDP protocol, will HTTP/3 become the next big DDoS reflection attack?

后续看了一些文章,在首屏启动的直播场景,理论功效还是比较大.


行为树

地址

熟悉我的都知道,面对性能优化与ReDesign, 我个人是比较热衷于升级数据结构的

黑板
考虑到节点之间或者行为树之间的输入输出, 引入了 黑板(blackboard) 的概念, 分为几种:

global: 所有行为树共享
per-tree: 一棵行为树的黑板, 该树的所有节点共享
per-node: 单个节点的黑板, 主要用于节点

上面这个抽象着实不干净.


帧率与卡顿

地址

虽然我没有意向转向游戏行业,不过这个视频对于游戏玩家来说真的是挺有必要的.

面向渲染
    - 逐帧 - 动作游戏
    - 每秒 - 网络游戏

> 天涯明月刀 dnf 这种网络动作游戏呢? 显然也是后者了.  
> 街霸 5 联机就不一定了. 

有关连接池大小

地址

最近思考工作的一些问题重新把光的这篇翻出来读了一下

The question of how many more also depends on the disk subsystem, because newer SSD drives do not have a "seek time" cost or rotational factors to deal with. Don't be tricked into thinking, "SSDs are faster and therefore I can have more threads". That is exactly 180 degrees backwards. Faster, no seeks, no rotational delays means less blocking and therefore fewer threads [closer to core count] will perform better than more threads. More threads only perform better when blocking creates opportunities for executing.

当然也有

Pool sizing is ultimately very specific to deployments.

For example, systems with a mix of long running transactions and very short transactions are generally the most difficult to tune with any connection pool. In those cases, creating two pool instances can work well (eg. one for long-running jobs, another for "realtime" queries).

In systems with primarily long running transactions, there is often an "outside" constraint on the number of connections needed -- such as a job execution queue that only allows a certain number of jobs to run at once. In these cases, the job queue size should be "right-sized" to match the pool (rather than the other way around).

DIY Single Sign-On for SSH

地址

这个就很有意思了..

文中提到的step也很有意思

brew install step

一些零散的记忆

行文至此突然想起来有人以前提过想以MitM的思路去做 Mock 平台…


done.