Hacker News new | ask | show | jobs
by mbreese 1662 days ago
I had a bonded DSL line for a while when I lived in the Bay Area with service coming from two different twisted pair lines. One line was consistently ~18Mbps, the other barely 3-5. It was pretty clear that one of the pairs was good and the other was broken somewhere alone the way. The lines were all in a bundle, with no way to discern what was what any individual strand was in the bundle (or which was broken or shorted). No one had any motivation to find the break and repair it. And because the line was technically “working”, ATT wouldn’t move it to a different pair. Sonic was the ISP with ATT handling the physical lines.

Still amazing that it worked at all.

1 comments

You can do time domain reflectometry to find out where the breaks, sharp corners and reflections are in the cable.

Some modems have special debug modes in that can do this too - then you get to know exactly how many meters along the wire the break is. When you get close, you can hook a resistor to the line and rerun the test and it'll tell you how many meters forward or back you need to go to find the issue.

Pretty easy to track issues down that way.

Yeah, this was in a neighborhood bundle with many… many lines together. No one was going to dig into that to find which line a mouse or squirrel had chewed the insulation off of. I remember running a few diagnostics, but as it was “working”, no one was going to try to fix it. The ISP couldn’t even convince ATT to move the bad line to a different pair. Which was sad. I was supposed to be in the 40Mbps range, but could only get ~18-20. Also — this was enough bandwidth for us at the time, so I just ran off of the single line and was good. Given some of the dsl horror stories, we weren’t too and off.