I.What a GPX track really says about you
Taken on its own, a .gpx file looks like an innocuous sequence of numbers: coordinates, elevations, timestamps. But unfolded, that sequence says a lot. It says where you were. It says when you go running, and when you come home. It says where you live (the point every track starts and ends). It says where you do your shopping, who your training partners are, where you spend your holidays, which days you work from home. One track is little. Fifty tracks is a biography.
On top of that come the physiological numbers when your watch records them — heart rate, pace, climbing speed. Enough to infer your fitness level, your approximate age, your daily rhythm. Not formal medical data, but something close to it.
II.The classic model: upload, account, leaderboard
The vast majority of GPX analysis tools follow the same pattern. You create an account. You upload your track to their server. The server parses it, stores it, analyses it, and sends back a page showing the figures. The track stays on the server — for their catalogue, their leaderboards, their segments, their algorithm training, their ad targeting, their resale to partners under each privacy policy.
This model makes sense for Strava or Garmin Connect, whose value comes precisely from aggregation: leaderboards, shared segments, friend tracking. Without aggregation, no social platform; the cost is the price of a service. Our comparison of Strava segments vs free isolation details the usage-side friction; our honest comparative places each tool back in its role.
III.The silent cost of uploading
But the model has side effects that aren't always highlighted.
- The track persists. Once uploaded, it lives on servers. You close your account? Copies remain in backups, logs, exports to partners. Real deletion depends on each operator's policy, and is never instantaneous nor total.
- It can be exploited in aggregate. In 2018, Strava's public heatmap revealed the precise location of US military bases — simply because soldiers were running with the app on. Individual data, aggregated, becomes something else.
- It feeds an economic model. Depending on the service, your track contributes to algorithm training, to ad targeting, to the sale of anonymised profiles to advertisers. You're not the customer, you're the fuel.
- You have to create an account to read your own data. The asymmetry is striking: to analyse a file you recorded on your own wrist, you have to sign up for a third-party service. It has become normal — it remains strange.
An uploaded GPX track is a signed document: where you were, when, at what speed, with what heart. Multiplied by fifty outings, it's a biography.
IV.The other model: everything in the tab
There's another approach, technically possible for years but rarely applied: do all the work in the browser, without ever transmitting the file's contents. That's what GPXchunk does.
Concretely, you load your .gpx file by drag-and-drop. The browser opens it directly from your disk, reads it, parses it, and runs every operation inside the current tab — distance computation, elevation filtering, map rendering, elevation profile drawing, measure recomputation every time you adjust the segment. No network request carries the contents of your track. You close the tab, the track is gone from memory. No copy on a server. No log. No database. No account.
It's an architectural guarantee, not a promise of goodwill. The difference matters: a service that uploads your file but promises « we won't touch it » depends on the trust you grant it and on its remaining the same operator (acquisition, leak, change of policy). A service that never receives the file cannot betray it, by construction.
V.What does leave the browser: the honest qualifier
« Nothing leaves your browser » deserves a clarification. Three things do leave, and it's worth naming them.
Map tiles. The map is rendered from images served by OpenStreetMap. When you look at your track, your browser asks OSM for the tiles covering the displayed area — so your IP address visits their servers, briefly. OSM doesn't see your track; it sees that an IP requested a portion of the map. If even that bothers you, you can analyse your file offline, without the map (the elevation profile and every measure work without a connection).
Ad scripts. GPXchunk is free and funded by Google AdSense. Under your consent (or implicitly outside the EU, UK and Switzerland), AdSense and Google Analytics load their scripts to measure audience and display ads. They see that a visitor consulted the page, not the contents of your file. The distinction is sharp: they measure the site, not your data.
The page's own resources. Like any web page, the browser downloads the HTML, the stylesheets, the scripts. None of these requests carries information about your file — they serve to load the application, not to read your track.
VI.Privacy-first isn't paranoia
The « nothing leaves, nothing remains » logic isn't a fetish for the privacy-obsessed. It's an architectural choice that opens simple, real use cases.
You want to analyse an outing before deciding whether to publish it. You want to look at your figures without that consultation being recorded by a third-party service. You want to open a track a friend sent you without having to create an account to read it. You want to analyse on a hotel Wi-Fi without committing the file's contents to infrastructure you don't control. You want to understand why you cracked on Sunday's last climb without owing anyone an explanation. You want to understand your performance without entering an attention economy.
None of these cases is paranoid. They are normal, ordinary ways of wanting to read a file you own without registering it in someone's catalogue. That's the service GPXchunk provides, and that's all.