March 26, 2015

I am at a conference. It is day four. It is very interesting but it is also a mix of science, pseudoscience, fringe science and then a lot of encouragement to sell patients "neutraceuticals" and "medical food" out of my office. The last bit majorly pisses me off.

The last speaker yesterday did a talk titled "Exploring and Facilitating Challenges in the Journey of Change". Yes, it's that sort of conference. The physician said disingenuously that at first they "lost a lot of money" on the neutraceuticals "because we are health care providers. We don't care about money. We GAVE our patients the neutraceuticals and we TRUSTED them to pay." Uh, right. Shee-it. I left. That was enough of that.

It is an expensive conference. They said it was expensive because it is NOT supported by (evil) pharmaceutical companies. However, it's at a stupidly luxurious conference center, and it is very thoroughly supported by multiple neutraceutical companies, special labs and "medical food" companies. They have nice hour long breaks with snacks in the propaganda room to lure us in.

I am staying with a Great Uncle, who is very conservative, instead of at the stupidly luxurious place. I did manage to get into the pool yesterday, have avoided paying for parking, and went for a hike in the desert during lunch yesterday. I asked the conference staff table people why we are at a stupidly expensive place and said I found it offensive. They said that they can't get the "right" food, gluten and dairy free, unless they are at a Very Special Place. I think that is a lie. Also, the gluten and dairy free lunch the first day had wheat berries. I kind of thought they contain gluten. There are containers labelled cucumber-blackberry water in the propaganda room, with cucumbers and blueberries floating in them.

Anyhow, some of the science is good and I think some of it will make sense to patients. They are talking about spending time with patients and actually listening to them, which I thought was what we were supposed to do as allopathic doctors. My recent medical trip has revealed to me that almost none of the doctors I saw actually did listen. Two did. I am rare in that I want to hear the story.

The conference cheerleader folks keep telling us that we are the few, the special, the cutting edge, who will actually help patients. I think the best thing I can do is listen to patients and in my area, recommend that they plant and eat kale. And avoid pills, as few pills as possible, whether pharmaceutical or neutraceutical or natural or bloody unnatural as hell... I ain't seen no pill growing on a tree yet.

This conference seriously makes me want to quit medicine.

This week has been a massive pain in the ass. For those of you aspiring to be software developers, there's one aspect of it that isn't taught in school nearly well enough, and is something that is the driving force behind every release. Performance. Once the core aspects of a product are completed, performance becomes the most important thing you will focus on ad nauseum. Every new feature that gets added should either not hurt performance or improve it. And it is here that the ideals of academia get thrown out the window. There is the theoretical way to code something, and then there is the performance driven way to code something. Just wait until you see some of the crazy things that get added for the sake of gaining 1%...

Everything starts off at the theoretical ideal. Then performance regresses, with a week left before Project Managers start freaking out because we're behind schedule, and you're stuck doing run after run after run, tweaking something here, commenting out something there, build after build. Performance regression defects are the bane of my existence. Give me the core dump of a crashed process any day - at least there's a reason it crashed; after going through the stack, and the memory regions, and the assembly code, you will eventually determine why it crashed. With performance there is no standard approach besides obtaining profiles and trying to make some sense out of the data it provides. To add to that, you have to run the thing multiple times in order to get any kind of confidence that what you're measuring is the stable value and not just random fluctuations. You also have to try and normalize the machine you're running on to the best of your ability. This means pinning processes to CPUs, having multiple machines, adding physical network links between machines, etc.

It is a black art that I have not come even remotely close to mastering. And it is a giant pain in the ass. But as much as I hate it, I understand that it is absolutely necessary. If there are two competing products that do the same work, wouldn't you want the one that performs better?