Friday, September 15, 2006

familiar strangers

My stay here in Fresnaye is coming to an end. Tomorrow I head up to Pretoria for a couple of weeks of teaching, training, helping out, and having fun. This Saturday the Pretoria group is going to a lake for some fun on the water. I'll be back at altitude - they're at 5,000 ft or so, I think.

I read once the term "familiar strangers." Though the power of Google I see it's originates from Stanley Milgram in 1972 and that my understanding of it was wrong. Quoting one page: "By definition a Familiar Stranger (1) must be observed, (2) repeatedly, and (3) without any interaction." I've been in this area for about 6 weeks. Main Road is a few blocks away, with many stores, shops and restaurants. It's reminds me of television shows I saw growing up, most specifically reruns of "The Streets of San Francisco." Cape Town and San Francisco have much in common. Despite the beach it doesn't have a Miami feel; more sepia than pastel.

This is part of the city proper, constrained by geography between Lion's Head and the Atlantic. People build up not out, with many 3, 4 and 5 story buildings. Short enough to not need elevators, tall enough to get density to support 3 grocery stores within easy walking distance, 4 laundry places, 4-5 internet cafes, tool rental, mobile phone service, seamstresses and tailors, clothing shops, a toy store, a pasta place - which is not a restaurant, and more.

The streets are made for people. The main street is mixed use zoning with housing above shops, though off main it's almost all residential. It's a street supporting providing services to the people who live here. There's some street parking but not much. This is still spring and the summer crowds will certainly change that. Go out into the suburbs and of course that changes, with large stores and malls surrounded by free parking. Here you pay. Every block has a parking attendent, employed by the city I think, to take your money and give you a receipt. More formal than the parking guys more often present in the evenings and nights. I hate parallel parking on the left with a right-hand drive car and so totally appreciate their help.

I go to Que Pasa on Thursdays. It's always the same parking guy there. He recognizes me and I him. I walk down the street here and see the woman who works at Lolly's, a nearby cafe, or the woman at the laundry, the man behind the counter at the internet cafe. I assumed these were familiar strangers because I know little about them. I don't know their names, or anything else other than the limited role from their jobs.

But it seems they are not strangers enough as I've interacted with them, while Milgram (according to secondary sources) uses that term for a "relationship in which both parties agree to mutually ignore each other, without any implications of hostility." I have not ignored these people.

There are others who are true familar strangers. There's a guy who lives across the street. On the street. He used to sleep in the bus stop shelter but a few weeks ago it disappeared. It was there one day and gone the next. He slept on the slab for a while but not he sleeps in the green space off the sidewalk. During the day he mostly talks loudly to himself. Like I said, many things here remind me of San Francisco. He and I have never interacted but I would be (slightly) concerned if I didn't see him around.

The "implications of hostility" reminds me of an account in Diamond's "Guns, Germs and Steel" where in Papua New Guinea when stranger met they would work how how they were related, even if distantly. Elsewise there would be the cultural obligation to fight. Given the world's long history of distrusting strangers, I am amazed that people can now travel widely and not fear, say, being stoned for not being one of The People.

On Tuesday evening I went to dinner with Åsa, whom I know from salsa. She's Swedish (in case you couldn't tell) and I practiced some with her. It reminded me of just how much more I have to go. I know enough to make people think I know the language when I'm really just doing a lot of inferencing. A friend of her's - phonetically "Lynn" - came at the end before heading over to salsa. Her friend has been here for 8 years and has the South African accent down pat. Not just language use like "hey?" for "sorry?" (as in "sorry; could you say that again?") but really all throughout the accent. Quite impressive. Åsa said she first thought my Swedish pronounciation had a Norwegian accent and that she doesn't hear (much?) of an American one.

Wednesday was dinner with Sylvia from tango. She's one of my favorite dance partners here because with her I can be extemporaneous and let everything just flow. With some others I have to think harder about what I'm doing. For example, with beginning dancers I try to be very clear and not do fancy things. Even with more advanced dancers some just don't like leans and dips and the other salsa-influenced things I can add.

Wednesday day I went over to NBN central and to SANBI. I got reimbursed for the flight down here. It's much more of a hassle for them to wire money to a foreign bank account to pay for a plane ticket. I would have to wait until I leave so they have proof that I used the full ticket, or something like that. They don't need that if I'm paid in cash. Now the largest common bill here is R100 (about $15) and the flight was about US$1,600. That's a stack of money. Guess I'll but some new clothes before going back to Sweden, though perhaps not as they don't really do winter clothes here. New glasses perhaps? Shoes? But I have a rather small luggage allowance at 22kg so I can't buy much.

I spent some of the last couple of days working on "Martel v2". Martel was a parser generator I wrote some years back for the Biopython project. It was supposed to simplify parsing the types of file formats often found in bioinformatics. In that respect it failed. I as the Martel author still found it easier to write parsers by hand than use Martel. It did work, but it was hard to use and debug. Martel 2 uses several different approachs to address and hopefully fix some of the problems in Martel 1. It's getting there but there's probably another week of development time before it's usable by anyone else.

When I dropped by central I talked with Paul some about my new approaches. He's the Python lecturer for this year, which is what I was last year. He's here for all 2+ months of the class, which is quite impressive. I got rather burned out after 5 weeks, though I did also teach usability and chemical informatics, and offered some help with basic probablities. As an interesting one, one of the students is Moslem and she didn't know the details of a deck of cards. My mother's family was pretty strict and they also didn't play cards nor were allowed to go to the movies. My dad's parents had a Rook set, which I've since learned was more common in religious households because it's different from a normal regular deck. Though I think you could still use a Rook deck to play poker, with a bit of mental renaming.

I then went upstairs to the NBN to talk with Heikki about DAS2. I'm a bit stymied with DAS2 development because I don't have good domain expertise. I have some GFF3 data sets which I can convert to DAS2 but there are fields I don't know how to convert properly. I also want some more complex data sets. The DAS2 feature model is more complex than GFF3 and I want something which can stress various parts of the spec. Vlad has a regulatory dataset which might be appropriate, and Heikki filled me in a bit about some of the nuances of SNP and haplotype data sets.

In explaining DAS to him I found I need to rethink how I explain DAS2 to others. It's really several parts. There's the reference sequence, which is pretty well understood. There's the types document, which I had to stress is not a type system. And there are the features, which are hierchical components located on subranges of the features. Those together are the core of the DAS data model.

The expected default visualization for this is a set of tracks, one data type (or ontology type) per track, with the regions denoted. This is an implicit part of the DAS spec but should be mentioned explicitly.

The last is the search interfaces. It's a simple filter-style interface with a required set of filters and some extensibility built-in. For more complex queries there's yet another way to extend the system.

Too configurable means that no one supports everything and interoperability suffers. I think we're pretty close to the right balance, but I want to test it out, and come up with a set of recommended use-practices for those converting from other data sets into DAS.

There's a lot of work left to do, especially given the other projects I've been working on and the 3 weeks of teaching I'll be doing over the next month.

No comments: