mixRadio — Internet Radio

2010. Before the Mainstream

“Internet radio” sounds routine now. In 2010 — it was bravado. Icecast, playlists, narrow pipes, office listeners — pure startup vibes.

I built my own station and a similar system for vrespecte. Streaming, web panel, stats — with a main enemy: memory that leaked like sand through a sieve.

Business Insights

• Uptime = trust and lower ops cost: fewer crashes → fewer pager duties.
• The cheapest optimization is leak elimination.
• Reuse (radio as monetization in vrespecte) increases returns on R&D.

The Problem

After 6–8 hours the server died. RAM ballooned, OOM killer did its job. The hunch — “something isn’t freed.” In reality — classic leak from unclosed resources.

The Fix (honestly short)

• Reproduced and measured memory growth.
• Found unclosed file descriptors at the hotspots.
• Switched to explicit resource management (context managers) and tightened caches.

The result was both funny and beautiful: after the fix, the system ran for years. Literally. Last time I checked the uptime was near eight years.

What It Taught Me

• Respect explicit resource management (“explicit is better than implicit”).
• Keep caches on a leash.
• Python services can be “like a tank” if you sweat the details.