Blog: Retaining Cyan

Retaining Cyan
Retaining Cyan
Date: 2018-Feb-08 00:19:10 EST

I've been a bit annoyed recently at the worry of database handle leaks in Go code. defer might not be enough. I'm hoping my most recent patch will help; as of the previous one, it looked like after about a day the server exhausts its supply. I wish the scoping rules for Go were saner. Go's kinda deceptive in that while it's more dynamic than C, it's not quite dynamic enough to be really civilised. The first glance is good, but in the long run it's meh. Title of this entry relates to this; when the handles are exhausted, the software's database needs are not met and it can't supply CSS to colour its entry screen cyan.

Today was full of meetings. Which is pretty cool when they're all about science, but rough when one lacks the background for some of that science. The machine learning reading group keeps challenging me, but the idea of turning Bellman equations into code led me down some interesting routes, among them thinking about how memoisation is made complicated by distributed systems. I think memoising is pretty much essential to making these things work well, but as a key idea in it is keeping access times for results faster than computation, we'd immediately run into the challenge of how to do that in a distributed system. Entirely separate per-worker caches? Doable, but that'd waste a fair bit of computation. Regional with mild access penalties? Workable. Global? Those computations had better be expensive to make that worth it. There is some smarts that one could work in there ; find some way to bias job distribution along lines where the caches are more likely to be effective. I've been thinking about ways to do that.

A few meetings back I started thinking about the relationship between fourier transforms (and variants on them) and factorisation of numbers. It feels strange that I think so often about factorisation and primes now; it never was a strong interest of mine before. Probably an effect of being around other science-y people again.

I don't place much stock in "brain training" companies, but this question and the nuanced discussion after it was pretty great at showing the distinction between narrow economic/mathematic understanding of a problem and bigger-picture views; "the best deal" for bulk buys of a product is in some senses fairly different from the best per-unit price, in that if it's something you don't need a lot of, storage (or discard) costs easily outweigh the transaction efficiency. A lot of people make this mistake in a lot of contexts.

I've been meaning for awhile to write about the evolution of how I understand political difference. I've come to think about politics as being abstractions on various spectra for positions on beliefs and styles of judgement, captured by tribes through identity and taboos. What this means is that people might naturally have a variety of beliefs on a lot of things, and each individual topic typically has a spectrum (or a space, if it's a complex topic) of positions possible. At least in theory people can have almost any permutation of beliefs, particularly but not only if there is no deeper philosophy involved in bringing one or another order to those beliefs and they don't generally try to make them consistent. Styles of judgement are also a big part of this, with a 3-polar model of issues capturing most variation. In this, one can:

  • neglect an issue - decide it is best handled in a hands-off way. When people are involved, we'd let them make their choice and deal with the consequences
  • nurture an issue - decide that offering resources, programmes, and positive support and possibly distractions are the best way to handle the issue.
  • discipline an issue - decide that penalising or blocking some behaviour on the issue is the best way to handle it
I think we often apply two to three of these approaches to issues. Finally, because most people don't like standing alone on issues and they find strength and comfort in holding them together, identities turn into tribes, and tribes lock off consideration on some issues by establishing and enforcing taboos. They often extend beyond the tribe itself, either through accident or through capturing of laws and rules in organisations. Tribes act to reduce variability on issues.

Still trying to get Tortfeasor to eat more food. I keep worrying about his rate of decline. Particularly bothersome is that for a long time he's occasionally walked around and theatrically plopped onto his side. I always took it as him being silly. His recent balance/strength issues and how he does that more nowadays recolours that past event in ways that make me worry I've been a neglectful cat-dad. The fading hours of a longtime companion are sad times.