Gesamtlänge aller Episoden: 96 days 13 hours 34 minutes
In this talk, Andor Penzes shows how to combine emerging technologies of functional programming. Idris is a dependently typed programming language, which makes it as a sharp tool in the programmer’s hand. A very good fit to modeling complex domains with high accuracy. Idris has a plugin architecture for the code generation, which allows engineers to implement code generations for even exotic technologies. Penzes' exotic backend is the code generator of the Glasgow Haskell Compiler...
QubesOS is probably the most secure operating system right now, recommended by security professionals and human rights activists. Security issues in operating systems are often related to networking or hardware. By separating networking, private data and hardware drivers into different virtual machines, QubesOS helps to isolate attacks. In case of a breach, attackers are confined to a virtual machine (=Qube), for example the networking Qube...
At the boundaries of our applications, a lot of bad things can happen. Users supply all kinds of invalid or malicious input, other services may or may not return what their documentation promises, little Bobby Tables is sneaking around the corner...
Reacts deklaratives Programmiermodel ist zumindest seit Hooks elegant und produktiv, führt jedoch leicht zu langsamen Anwendungen. So ist die React Weiterentwicklung (Fiber, Concurrent Mode) der letzten Jahre vor allem durch Performanceverbesserung motiviert. Was sind die typischen Probleme, wie kann man sie diagnostizieren, welche Optionen hat man? about this event: https://bobkonf.de/2021/schmalhofer.html
Servant is a well-known library in the Haskell ecosystem to develop REST services. Mu is much newer, and tries to fill the gap of gRPC and GraphQL services. Both libraries have in common their use of type level techniques to describe the structure of the services. This talk dives into that design corner, and explores the different decisions taken by both teams. about this event: https://bobkonf.de/2021/serrano.html
Data science often involves composing individual tasks to complex workflows. A popular tool to quickly implement and check individual tasks is using UNIX pipes. The philosophy behind pipes is implementing small programs that “do one thing well” and compose them via pipes. The composition should adhere to one universal interface, passing strings both as input and output. While the interface is simple to understand and design for, however, bigger tasks quickly become unwieldy...
Produktentwicklung ist geprägt von kurzen Iterationszyklen und wird darum gerne agil betrieben, wie in meinem Team. Dort wurde die vorhandene unstrukturierte Architekturdokumentation in das arc42-Template überführt und in einem Wiki abgelegt. Das Template hat den Vorteil leichtgewichtig und in sehr vielen Formaten - darunter Word, LaTeX und AsciiDoc - verfügbar zu sein...
Modern processors support Single Instruction Multiple Data techniques since 1997 when intel introduced MMX. Every modern CPU has physical registers for performing vectorized SIMD operations on up to 512 bits with a single instruction. There is a part of the CPU which is most of the time only used by compiler optimizations invisible and unbeknownst to users or by low-level systems programmers comfortable with assembly and processor architecture...
This example-driven talk introduces and explores “Higher Kinded Datatypes” (HKDs), a novel approach to parameterizing data types which allows greater flexibility, re-use, and improved type-safety when performing common tasks on data. While many data types are parameterized over the values in the type; HKDs are generalized over a container type which wraps each field...