26 Jul 2002 Bram   » (Master)

Load Testing

BitTorrent continues to shrug off every load test we throw at it. I thought a 700 meg file would get more concurrent downloaders, but instead it's just resulted in more people cancelling before download completes. The 140 meg file had an 80% completion rate, the new one is more like 35%.

I'd really like to get a load test with 200+ concurrent downloaders since that makes the connectivity graph more sparse than densely connected, but it's looking very difficult to make that happen. Earlier versions got that kind of load and didn't have artifacts related to large networks, so I may just give up and declare it final without another big deployment. The network is a manually constructed to be random, rather than following a power law, so it's unlikely for anything weird to happen. Still, another test would be nice.

In any case, recent tests have gone unbelievably smoothly. The only artifact of note is that when the load suddenly jumps two orders of magnitude it takes a few minutes to get warmed up, which is hardly alarming or even surprising.

Diary Ratings

Current diary ratings seem to conflate two concepts - your estimated quality of a diary, and your confidence in that estimation. That said, I'm impressed with how well the numbers reflect my own own opinions.

Diaries which talk about the author's life seem to rate worse (at least for me) than ones which emphasize technical topics. I wonder if this is a stylistic difference of opinion, and whether other peoples's estimated values have the same phenomenon or possibly even the opposite one.

Ugly Hacks

wardv just got me to throw up on my keyboard. That's more a possible new source of security problems than a useful technique. I wonder if parseargs has buffer overflows.

Picking on Wolfram

Wolfram presents a turing machine which he claims is universal because it emulates rule 110. That assertion is just plain wrong. The turing machine in question doesn't halt in response to a computation ending, it merely stabilizes, a clear violation of the ground rules for how turing machine encodings work. It also requires an infinite pattern on the starting tape, rather than a finite number of cells representing the problem encoding and the rest in a uniform 'blank' state. This is an equally clear, although less egregious, violation.

Latest blog entries     Older blog entries

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!