Weekend long read suggestions 2020/10

• US court fully legalized website scraping and technically prohibited it

  Coverage of decision which took part of the LinkedIn vs hiQ Labs. The appeals court also upheld a lower court ruling that prohibits LinkedIn from interfering with hiQ’s web scraping of its site. This fundamentally changes the balance of power in dealing with such cases in the future.

• The dark side of .io: How the U.K. is making web domain profits from a shady Cold War land deal

  Came to this (older) piece via recent decision by draw.io to rename to diagrams.net a they are not happy with the sate of .io TLD.

Software Development

• Porting a JavaScript App to WebAssembly with Rust

  3 part series mapping an exercise of rewriting React+Redux application written in JavaScript to WebAssembly (WASM) with Rust.

• Breaking Out of Ecto Schemas

  Super short article about stepping out from Ecto.Schema closer to raw SQL with Ecto. Neath trick is to use Map in your select: statement, so Ecto will return (list of) Map instead of a list of lists.

• Cheap tricks for high-performance Rust

  Pascal shares some of the simple tricks to speed up your Rust programs without really changing the source. Hints like properly setting your target architecture, alternative allocator, release profiles and more.

• When Bloom filters don’t bloom

  Marek needed to deduplicate large list of IP addresses, so he set sail on the journey of getting better then sort | unique. He shares some lessons learned about random memory access latency, power of cache friendly data structures and Bloom Filters and finally “just” hash table.

• Moving a method from struct impl to trait causes performance degradation

  On very similar note as a previous one – code alignment having a significant impact on the performance.

• Starter project for Flutter plugins willing to access native and synchronous rust code using FFI

  Flutter meets Rust, wow.

psql

• On recursive queries

  I never really give a recursion though in the context of SQL…

• PGTune

  pgtune on web. Simple site to give you psql configuration to start with for different use-cases and server configuration.

Machine Learning

• BERT Related Papers

  List of various papers related to BERT (Bidirectional Encoder Representations from Transformers). BERT was released by Google as part of their NLP research. But researchers are stepping forward and you can find multi-modal applications as well.

• Exploring Transfer Learning with T5: the Text-To-Text Transfer Transformer

  New publication by Google at the NLP space. They have published a new model called Text-To-Text Transfer Transformer (T5). They have also open-sourced a new pre-training dataset, called the Colossal Clean Crawled Corpus (C4).

• Transformers are Graph Neural Networks

  Drawing parallels between Transformers (key component of BERT) and Graph Networks.