In a podcast by people from Adaptive Path they were talking about their failures and maybe even why failure was good. One of them said that failure is often a sign that you were trying new (and presumably, interesting) things. It occurred to me that success is a drag, if you succeed you now have a legacy app to maintain! Whereas if you fail, you can run on to the next interesting thing. My experience is that the only way to escape a successful, now legacy, app is to leave the company. And even that doesn’t always work. Over ten years after my first for-pay programming job, a summer job, my employer tracked me down and wanted help.
Archive for July, 2007
Wireless DHCP Oddity
Monday, July 16th, 2007Last night, the DHCP server at the hotel where we are currently living appeared to go on the fritz. All three access points are visible, but the DHCP lease isn’t renewed in spite of numerous tries. This continued all night and into this morning. I’ve seen this before at other WiFi hotspots. Checking with other users and rebooting into Windows just confirms the behavior, so I’ve always given up and working without Internet connectivity or gone somewhere else.
This morning I tried something different. Since the access points are up, what happens if I just manually set my wireless card’s IP address. The logs suggested that the last IP address was a private Class C address (192.168.X.X). It seemed unlikely a large commercial system would use Class C, Class B (172.X.X.X) seemed more likely. I did it anyway and immediately the DHCP server assigned a Class B address and I was up and connected.
Maybe this was just coincidence, but I wonder.
Mea Cupla, Bad Engineer – No Donut!
Monday, July 2nd, 2007I ought to know better by now. Measure, then cut. I have been working on Amethyst, an adaptive RSS reader using Bayesian classification to (try to) train a program to show me news in most interesting to least interesting order. I have tweaked and fiddled, etc. Last night I coerced DSPAM into coughing up the most/least interesting phrases that it is using to classify. For RSS feeds with a lot of low ranking stories, the significant words and phrases are bits of HTML tags, font names, etc. I.e., it is the formatting that is being graded, not the content. Extracting the content from the HTML is a non-trivial problem. I think just a list of words to ignore may be an easier solution.
If I had looked at the actual data, I would have saved myself a lot of wasted effort.