<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>API Design on Andrea Bozzo | Blog</title><link>https://andreabozzo.github.io/AndreaBozzo/blog/en/tags/api-design/</link><description>Recent content in API Design on Andrea Bozzo | Blog</description><generator>Hugo -- 0.146.0</generator><language>en-US</language><lastBuildDate>Tue, 20 Oct 2026 17:15:00 +0200</lastBuildDate><atom:link href="https://andreabozzo.github.io/AndreaBozzo/blog/en/tags/api-design/index.xml" rel="self" type="application/rss+xml"/><item><title>Where Should the Library Stop?</title><link>https://andreabozzo.github.io/AndreaBozzo/blog/en/posts/where-should-the-library-stop-blog/</link><pubDate>Tue, 20 Oct 2026 17:15:00 +0200</pubDate><guid>https://andreabozzo.github.io/AndreaBozzo/blog/en/posts/where-should-the-library-stop-blog/</guid><description>I built a small dlt + DuckDB pipeline that uses my own data profiler as a dependency, and wrote down every line of glue it needed. Seven gaps in the library I maintain, four in the one I don&amp;#39;t, none in DuckDB. On what a caller has to own, what the library should add, and the one feature it should refuse.</description></item></channel></rss>