99% of Developers Don't Get TCP/IP

### Transcript {ts:0} 99% of developers don't get TCP IP. You hit play on the first episode of Squid {ts:5} Game season 3. You push a oneline hotfix that accidentally takes down your entire team's prod database. You drop the this {ts:12} is fine meme into your Instagram gossip group chat. … And why is the {ts:45} future of the web, HTTP3, being built on UDP instead of the ultra reliable TCP? ... These {ts:210} protocols almost always use TCP underneath to guarantee reliability except for some DNS queries and {ts:217} real-time apps that use UDP. Next, we have the transport layer. The transport layer is where the transmission control {ts:223} protocol or TCP resides and it solves one of the internet's biggest challenges. … It then sends these segments and waits for acknowledgements or acts from the {ts:247} receiver. If a segment is lost or damaged, TCP detects the problem, usually through missing acknowledgements {ts:253} or corrupted check sums and automatically retransmits the affected segment. This process is called positive {ts:260} acknowledgement with retransmission or PR. To avoid overwhelming the receiver, TCP uses flow control via the sliding {ts:267} window protocol, which allows the sender to send multiple segments before requiring an acknowledgement, but within {ts:273} a limit that the receiver can handle. TCB also adjusts its sending rate based on network congestion. These monitor {ts:279} signs of congestion such as dropped packets or increased roundtrip times and throttle transmission rates when needed {ts:285} to avoid flooding the network. Because of these mechanisms, TCP is considered connectionoriented. Before any data is {ts:292} exchanged, a three-way handshake occurs. … For applications where speed is more important than reliability like video {ts:326} calls on FaceTime, online games or live streams on Twitch, TCB can be too slow. In such cases, a different transport {ts:333} layer protocol called UDP or user datagramgram protocol is used. UDP is connectionless and does not guarantee … But that very strength the stable singular connection can become a {ts:376} massive weakness. Imagine you're not just loading one web page. Imagine you're a developer building a service {ts:381} that needs to gather public data from thousands of web pages. If you try to open thousands of these TCB connections {ts:386} from your single server, the target's website's firewall will see it instantly. Your server's IP address gets {ts:391} blacklisted and your project is dead in the water. This is the fundamental challenge of large-scale data {ts:396} collection. And solving that exact problem is why I'm going to introduce you to Dakota. … {ts:492} It abstracts away the fragility of individual websites, handling bot detection and site changes for you. This {ts:498} means you can build more resilient data pipelines that require less maintenance, ensuring your data links and vector {ts:503} databases are constantly fed with fresh, high-quality data for training and rag systems. … And for software engineers and QA teams, flaky end-to-end {ts:539} tests are a nightmare. Stop debugging test failures caused by your CI/CD runner's IP getting rate limited or {ts:545} banned by a thirdparty API you integrate with. By routing your Playright or Selenium tests through Dakota's network,