Pyright, Automation, Routing, Networking, Location codes
- Tested pyright, static Python type checker. One thing annoyed me, it's written in TypeScript and requires Nodejs. And then they say it's awesome, because it doesn't require Python. Wait what? Afaik, writing Python tools in Python would make perfect sense. If every tool uses "something else" it sounds just horrible. I wrote Microsoft SQL Server (TSQL) syntax checker in Python, and it requires PostgreSQL for check database. Nah, just kidding. But you can see the point.
- PowerShell system automation scripts. Deploy and configure Virtual Windows Hosts completely automatically, including all required settings and software. That's a lot of things to fine tune and figuring out how to get everything done. As mentioned I'm not a PowerShell ace. But all it takes is patience, lot's of reading and testing and bit of pain while learning new things. After everything is automated and working, it's totally awesome feeling to get things done completely automatically. I'm really happy to see all this stuff working perfectly. Even if there's quite a learning curve.
- At some point I might convert some of those ultra lame Python scripts which call PowerShell, into pure PowerShell. Especially in cases, where there's not that much logic or data structures to be stored.
- It's so nice to wake up in the middle of the night, because someone has screwed up routing. Now it seems that UpCloud - Elisa (Helsinki, Finland - Helsinki, Finland) traffic (IPv4 + IPv6) is looping via Stockholm, Sweden and giving nice 16 ms extra boost to latency. Which in this case is around 1600% compared to normal latency. Now Helsinki - Helsinki connections are nearly as slow as Helsinki - Frankfurt. Because normal latency is just below 1 ms. - Let's see if their NOC is efficient. A friend promised me a good meal, if they can resolve issue like this under two hours. - And fail. - Shortly after that they came back with report, which indicates that Elisa has chosen to change the routing for some unknown reason.
- After the active edge protection was turned on. It seems that the hosts which earlier lacked proper protection are attacked at much higher rates and from many more individual IP addresses than the few hosts which had old the non networked independent deference system installed. Also any new hosts added to the system see much lower attack volumes than the older hosts which lacked protection. Which just indicates what's expected, the soft targets which lack defense mechanisms are hammered hard all night long. Systems and networks which actively take action and block traffic, send a clear message to the attacker that this isn't going to work, so you can just as well stop right now. Of course this is no news to anyone, that's exactly the logic I would also apply when attacking systems.
- Played with WireGuard, which is a fast modern and secure VPN technology. Yup seems to be working. Of course tunneling between two Linux servers for fun. I personally like WireGuard design approach. Do things in a simple and reliable way. Without making things ultra complex, bloated with huge overhead and tons of legacy code included.
- Simple location codes for areas without addresses? Plus Codes which actually utilize Open Location Codes (OLC). Well, I've been writing so much different kind of mapping and encoding code. That this is just yet another such issue. Nothing new here. Which can be solved in many many several ways like S2 or H3. In some of my projects we've just used straight out coordinates. Because the data set wasn't too large, it was efficient enough. No need to convert those into "block" addresses or something similar for more efficient indexing. Yet when designing the system there was reservation for block based addressing. Where around 100 x 100 meters block would have been used globally, allowing faster data retrieval and weighting based on location.