It lead me to Conover Plastics, whose home page claims,

Conover Plastics, Inc. is the leading manufacturer and supplier of sock hangers. With over 30 years of experience making sock hangers, Conover has a great assortment of sock hangers. We have many standard sock hangers in stock - ready to ship the same day we receive your order.

Every once in a while, I get a glimpse into a corner of business I don't usually see, and am amused and impressed by the complexity of human activity. Here's a product I've seen over and over, and never thought about where it came from. And it turns out that there's an entire company that makes nothing but those non-descript sock hangers. Not just that, I'm sure Conover has competitors, there may be an entire sub-industry of sock hanger makers, or at the very least, specialty injection molded plastics.

The enormity of industry is an amazing thing. The fractal-like splitting of products and services down into component parts; the supplying and demanding of those parts from firms created purely to provide them; the trade associations, brokers, warehouses, and middle-men; the conventions, skill sets, sub-cultures and specializations are fascinating to me.

I partake in these sorts of specializations in the software arena, and to think that there's that sort of unseen complexity in sock hangers is astounding.

Somewhere at work we have a 200-page catalog just for shipping materials and another 200-page catalog of supplies just for manufacturing departments. Somehow we even a thick catalog just for audio-visual systems for churches.

This splitting into finer and finer sub-categories of specialization shouldn't surprise me any more, but it still does, and I'm proud and impressed to see the human beehive of activity at work.

Cog does one thing well: find snippets of Python code in text files, and run them, capturing the output and splicing it into the file. It can be used for all sorts of code generation and preprocessing tasks.

The irony of Cog is that I wrote it back when I worked in C++ and needed to generate code. Now that I work in Python, I don't need it, or so I thought. But this winter when I was writing my Whirlwind Excursion through Python C Extensions, I wanted a way to create the slides and the web page from a single source, and also not have to copy and paste sample code between runnable files and the text source. Cog would have been a great way to do this. For example, Paver uses Cog for just this sort of job.

But it didn't even occur to me to reach for Cog, and I ended up hacking something together. So I have to try to remember Cog in the future.

I updated Cog because I got two requests in one week to update it to keep it working on newer Python implementations. I'm going to start porting projects to Python 3.0, and Cog may be the first, since it is relatively small and simple.

One thing social networks do very well is keep people connected together who may not have been connected at all before. It used to be that only those closest to you would know when your birthday was. Now entering it into your Facebook profile means that all sorts of people that you're just a little connected to know it.

Some people criticize social networks for replacing significant connection with trivial ones, that they're the fast food of relationships, edging out meaningful interactions with empty friendship calories. I find them instead to be a supporting structure: the information and connection they offer is a scaffolding on which to build, and can keep you connected with people that you otherwise wouldn't be able to stay in touch with.

I woke this morning to find I already had a dozen good wishes from people around the world, including at least one I had never communicated with before. What could be bad about that?

I've blogged about Bathsheba before, but I just spent at least an hour poring over these artworks, and following connections from Bulatov's photos on flickr, so I thought I would share again. Also, Bathsheba has some models in plastic which reduces the cost.

If you haven't been keeping up with the changes since coverage.py 2.x, there are plenty of them, including greatly improved speed and a nice HTML reporting feature.

¶   Cryptographic Right Answers: with the latest best practices for crypto problems

¶   TDD Anti-Patterns: a funny skewering of what people do wrong in unit tests

Any recommendations?

I like that the diagram isn't symmetric: the "learn to monetize" segment directs you to move those activities to the center, as does the "learn to do this better" segment. But when you are doing something well, and getting paid, but it's not something you want to do, the instructions aren't to learn to like it: they're to say no.

I find the labels intriguing. Two I didn't understand, and had to look them up: 1 is Legendre's constant and 4 uses a negative modular exponent. The only one I don't like is 3, because it isn't a number, it's a numeral. It isn't a value that evaluates to 3, it's a character entity that displays "3".