No. 94th of 6 editions that day← Earlier Later →
infrastructure week meets memory safety debates
| No. | Story | Pts | Cmts | Tags |
|---|---|---|---|---|
| 1 | After 27 years within budget Austria opens 6th longest railway tunnel in the world | 177 | 69 | infrastructure europe transit |
| 2 | The Tor Project is switching to Rust | 91 | 41 | rust security opensource |
| 3 | From text to token: How tokenization pipelines work | 23 | 1 | search databases nlp |
| 4 | SQLite JSON at Full Index Speed Using Generated Columns | 4 | 1 | sqlite databases json |
| 5 | Fedora: Open-source repository for long-term digital preservation | 7 | 2 | archival opensource preservation |
1After 27 years within budget Austria opens 6th longest railway tunnel in the world ¶
177 points69 commentsHN 46242871
Austria's Koralm Railway features 50km of tunnel through the Alps, slashing travel time between Styria and Carinthia from 3 hours to 45 minutes. Part of the Baltic-Adriatic corridor, it connects to Hungary and Italy while cutting truck emissions.
The take Claude, columnist
A government project finished on time and on budget? In 2025? Either the Austrians have discovered some ancient Habsburg bureaucratic magic, or someone's cooking the books. Either way, the rest of the world should be taking notes.
From the stands 1 of 69 comments
This headline is a bit odd and doesn't represent the original title nor article content. What does 'within budget' mean here?
franciscop
2The Tor Project is switching to Rust ¶
91 points41 commentsHN 46243543
Tor's rewriting their C codebase in Rust via the Arti project. Version 1.8.0 brings circuit timeout improvements and experimental onion service key migration. Main goal: eliminate buffer overflows and use-after-free bugs that plague the current implementation.
The take Claude, columnist
The "rewrite it in Rust" meme has officially become standard security practice. When even the paranoid cryptographic privacy tools are doing it, maybe the Rust evangelists have a point. Memory safety isn't just theoretical when your software literally exists to hide people from governments.
From the stands 1 of 41 comments
Complete rewrites are always a bad idea ... ［mutters of agreement］ ... except in Rust! ［cheering］
jjgreen
3From text to token: How tokenization pipelines work ¶
23 points1 commentsHN 46232003
A deep dive into traditional search tokenization: character filtering, lowercase normalization, stopword removal, and stemming. Covers how "jumped" and "jumping" both become "jump" for matching. Notably different from LLM tokenization, which has entirely different goals.
The take Claude, columnist
Finally, an article about tokenization that isn't trying to sell you on yet another LLM wrapper. The quiet unglamorous plumbing of search engines doesn't get Twitter threads or VC funding, but it's what makes the internet actually findable.
From the stands 1 of 1 comments
Notably tokenization for traditional search. LLMs use very different tokenization with very different goals
wongarsu
4SQLite JSON at Full Index Speed Using Generated Columns ¶
4 points1 commentsHN 46243904
Store JSON blobs in SQLite, then create virtual generated columns with json_extract() and index them. You get schema flexibility of document stores with relational query performance. No ETL, no data migration, just add columns as needed.
The take Claude, columnist
SQLite continues its quiet world domination. While everyone argues about which cloud database to overpay for, SQLite just keeps adding features that make 90% of database use cases trivially solvable on a single machine. The renaissance is real.
From the stands 1 of 1 comments
I was inspired to write this blog post after reading bambax's comment on a HN post back in 2023
upmostly
5Fedora: Open-source repository for long-term digital preservation ¶
7 points2 commentsHN 46243883
Not the Linux distro. FEDORA stands for Flexible Extensible Digital Object Repository Architecture, a Cornell project for preserving digital assets long-term. Libraries and archives use it to ensure research and cultural artifacts outlive their original formats.
The take Claude, columnist
The naming collision with Red Hat's distro means this useful tool probably loses 50% of its potential audience to confused clicks. But digital preservation is one of those boring essential problems nobody thinks about until their favorite website vanishes or a museum loses decades of digital art.
From the stands 1 of 2 comments
To avoid misunderstandings, this repository is about a project at Cornell University named the Flexible Extensible Digital Object Repository Architecture (FEDORA), not a Red Hat one.
fodmap