A common recurring pattern I’ve observed is generating ingress hostnames from app names:<\/p>\n
app<\/span>: <\/span>foo<\/span>\n<\/span><\/span><\/span><\/span>hostname<\/span>: <\/span>${app}.<cluster-domain><\/span>\n<\/span><\/span><\/span><\/code><\/pre><\/div>It guarantees uniqueness and simplifies setup. For internal services or\nephemeral environments, this is often perfectly fine.<\/p>\n
But when this pattern becomes the default<\/em> for externally visible\nendpoints, it starts to break down — because it fundamentally\nmisunderstands what a hostname represents in that context.<\/p>\nA hostname is not just a unique identifier. It’s a contract — with\nbrowsers, with TLS, with DNS, and with your users. By reducing it to a\nderived label, you lose control over things that matter in production.<\/p>"},{"title":"Making JSON Awaitable","link":"https:\/\/fhalde.github.io\/posts\/jsontap\/","pubDate":"Sun, 01 Mar 2026 22:05:50 +0100","guid":"https:\/\/fhalde.github.io\/posts\/jsontap\/","description":"
When working with LLMs for building Agentic Apps, one thing quickly becomes obvious:<\/p>\n
\n- You want structured outputs.<\/li>\n
- The model streams tokens.<\/li>\n
- Your code waits for a finished JSON blob.<\/li>\n<\/ul>\n