Monday, November 24, 2008

Continuing Internet problems

I still have Internet problems, and I've seriously backdated this post to keep things in chronological order, so that I can post more in the right order if France Télécom's bumbling idiots ever fix their network. My apologies for the technical nature of this post, but the problem is very preoccupying at the moment, destroying my entire holiday “vacation” (in quotes because I don't actually get paid for days off, there just isn't any work at this time of year).

I've spent a few days looking at network traces. FT has an upstream problem that extends no lower than the DSLAM; the line and modems are good. About 50% of downstream packets are lost or dramatically delayed by the upstream network, particularly those of non-trivial size and those using HTTP protocol. This means that virtually all connections hang after a few seconds. Some packets never appear at all, others appear five minutes late, and many are in random order. The TCP/IP stack can handle random order, but not vast numbers of lost packets or packets that are so late that the application gives up on the connection.

So this essentially stops my Internet connection. FT and Orange (same thing, really) technical support is totally, inexpressibly incompetent, literally working from simple scripts that cover nothing except possible Windows problems and modem and line problems. Apparently FT thinks that its network is perfect and requires no surveillance or technical support, so anything that happens upstream of the line is a dead end if you try to complain about it. I've taken to publishing my measurements on the Internet, and I might even decide to print the data on handouts and give it to people in front of an Orange store so that they can see how France Télécom plans to shaft them (to me, 9 kbps out of 8000 advertised by the company is a pretty clear case of fraud).

Anyway, I managed to sneak this post onto my blog, hopefully the problem will be fixed sooner or later and I can return to regular Paris-related programming here.