About observer

What bitmagnet hears

Bitmagnet doesn't index a tracker. It joins the DHT, the peer-to-peer directory BitTorrent clients use to find each other, and records whatever passes by. That makes the catalogue a record of traffic the crawler happened to overhear, shaped entirely by what was being announced while it was listening. Anything missing may simply never have crossed its path.

What the DHT carries is mostly an info hash and a name. Turning that into a film, a series, or a book means matching it against a metadata record, and a lot of what bitmagnet hears never gets matched, because the name is too mangled, too generic, or describes something no catalogue has ever heard of. That unmatched share makes up most of what a DHT listener picks up, so this UI shows identified and unidentified results side by side instead of hiding either one.

Reading a row

Each row tells you first whether it was identified, through the typeface it uses. The two examples below aren't real rows and don't respond to clicks.

Identified

A metadata match exists, so the row leads with the title in IBM Plex Sans and its release year, with the raw filename underneath in IBM Plex Mono. Sans is how you can tell at a glance that some catalogue recognises this as a known work.

Unidentified

When bitmagnet tries to match the name against a metadata record and fails, the row shows the filename exactly as it came off the wire, in mono, labelled Unidentified. Mono marks it as a raw string bitmagnet couldn't interpret.

The gutter

The 3px amber bar down the left edge shows swarm size. Swarm sizes are heavily skewed. Most torrents sit in single digits while a few carry thousands, so the meter runs on a log scale and tops out at 1000 seeders. Anything past that just reads as full. A thin bar sitting right at the bottom means the swarm was measured and found dead, which is different from an empty track, where it was never measured at all. Hover any real row for the exact counts.

What the numbers don't say

A few of the numbers on this site aren't as literal as they look.

• A ~ before a facet count means it's an estimate. Once an aggregation runs past its time budget, bitmagnet switches to sampling instead of counting exactly, and the tilde stays there so the count doesn't look more exact than it is.
• The dashboard's first sightings figure can be larger than the whole catalogue, because it counts discovery events rather than torrents currently in it, and some of those get merged into one record or dropped later on.
• Only contentType, releaseYear, videoResolution and videoSource accept an "unidentified" filter value, because bitmagnet's schema only allows a null filter on those four. The other facets don't support it, so their filter lists never offer that option at all.

Rate limits

This site runs on a single bitmagnet instance, so every search you make costs it a real query. Ordinary use, such as typing, paging through results, or flipping facets, stays comfortably inside a normal visitor's budget, but that budget isn't unlimited.

Search more than ten times in a minute and the next one comes back asking for a quick check instead of an answer. That check is a short computation the browser solves on its own, without sending anything to a third party or requiring an account or cookie. Solve it and your search goes through right away. That only restores your own search budget, though. The hard limit underneath it stays fixed no matter how many of these you solve, so a script can't buy its way to more load than an ordinary visitor gets.

Creating an account asks for the same kind of check, once, for the same reason. It's the one free, unauthenticated action on the site, so it's the one worth making a script pay a little for.

Colophon

The catalogue and the crawler are bitmagnet; observer only reads from it.

Type is IBM Plex Sans and IBM Plex Mono. Sans and mono are how a row tells you whether its name was understood, so the two faces have to read as clearly different from each other even at 13px. Colour is sodium amber on deep navy. Amber is the interface's only chroma, showing up only on links, focus rings, and the swarm meter.

This page is static. It calls nothing, so it still renders when the crawler is unreachable.