grenouille.com

» see grenouille.com as it lived, on the wayback machine

~ the obit ~

Not believing the number your ISP prints on the invoice is an old habit of mine. It started with the 33.6k modem and survived every upgrade after that, through the TTNet years and into whatever the current contract claims. What grenouille did was turn that suspicion into a public institution, run by volunteers, in a country that was not mine. Its measurements only ever covered French lines, so for a connection in Ankara there was nothing in it except envy.

It began as a complaint. In the spring of 1999 a set of Wanadoo cable subscribers formed the Collectif des Câblés Wanadoo, because the service they were paying for was inconsistent and the operator's answer was always that the fault sat at their end. Rather than keep arguing, they wrote a small program that measured the line and started handing it around that November.

The site itself went up in November 2000 and published what it called la météo du net, the weather of the network. Every French ISP got a page. Every region got its own reading. By late 2006 the association counted around 280,000 registered members, more than 50,000 of whom were actually running the client, with roughly 25,000 people a day arriving just to look at the graphs.

~ the rap sheet ~

~ how it worked ~

The measurement was volunteer-powered in the literal sense. You installed a program that woke up about once an hour, pulled a test file off a dedicated server over FTP, and reported back what it had seen. The server side aggregated the samples, and the front page turned thousands of individual home connections into a chart of how one operator was behaving in one department that afternoon.

Because the association shipped free software instead of a product, there were four separate clients written by different people in different languages. Camlgrenouille ran headless on Linux and the BSDs, PyGrenouille was the cross-platform Python build with an actual window, vbGrenouille installed itself as a Windows service, and the Perl one had stopped being developed by 2014. All of them copyleft. That is a very particular kind of project and nothing quite like it exists now.

~ the fights ~

The numbers were argued about for the whole seventeen years. Pulling a file over FTP was a defensible choice in 2000 and a much worse one by the middle of the decade, once the traffic anyone cared about had moved to HTTP and the average home line had a wifi bridge sitting in the middle of it. The client tried to discard any sample taken while the connection was busy and did not always manage it.

None of that stopped grenouille from being the reference. When a French operator quietly degraded a peering link or throttled something, this was where the drop showed up first, in public, with a timestamp attached. An ISP could dispute the method. It could not dispute that fifty thousand of its own customers were watching the line.

~ the day it stopped ~

On 20 June 2017 the site posted that it had permanently ceased activity, and that was the entire notice. No reason given, no successor named, and the forums holding years of French broadband history were not kept online. People found out late, sometimes months late, because anyone still running the client in 2017 was not checking the front page every morning.

What went with it was a measurement nobody had a commercial reason to make. The speed tests that survive are run by companies selling something adjacent to the result, and they measure one moment on demand. This one measured continuously, from tens of thousands of ordinary lines, and answered to a non-profit association rather than an advertiser.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »