Saturday, February 24, 2007

Summary of Los Angeles meetings

During the past week, I attended two Smalltalk meetings in Los Angeles (as described in earlier posts and in the Smalltalk World Tour wiki page). I would like to offer a brief summary.

The attendees included people working on education, people working with robotics (both at a college level as well as using Squeak to teach kids), voice recognition researchers, college faculty, professionals and students. It was very stimulating and exciting to see how creative people successfully use Smalltalk to accomplish their goals.

Friday's presentation of A Pattern of Perception went really well and was received with enthusiasm. There were several members of SDSU's faculty in the room, and it was fantastic to see how much the content resonated with what they are doing. I am sure there is plenty of opportunity for further positive collaboration and exchange of ideas.

This is very good, I am anxious to have a second helping already :).

Thursday, February 15, 2007

And an additional invitation!!!

Who said one couldn't get something out of nothing? John Dougan suggested that we should organize a meeting for next week in the LA area. By all means let's do that as well! Here's the wiki page where we can organize.

This is good stuff, more of the same please!

Wednesday, February 14, 2007

SDSU: an updated invitation!

As you may know, a winter storm is ending its trip through the east coast. I got a phone call from the airline a while ago: my flight to the west coast was cancelled, and the next seat they had for me is on Sunday (at 6:30am, of course). Thus, I can't be at SDSU on the 16th, and the presentation has been canceled for this Friday.

I wish I had not had to write that, but I hope you understand I could do nothing about it. I even double checked the flights myself --- no dice. But no matter how bad things are, it is always possible to make something good out of them. Roger Whitney has been very kind, and we have rescheduled the presentation for Friday, February 23rd at 1pm in room GMCS-308.

However, I have more news. I also received a phone call from JP Morgan. Although the presentation had been formally approved before hand, their compliance department later determined that some additional process has to be completed before the presentation can be shown. The things that have to be done are quite reasonable and can be taken care of very quickly. However, a signature of a certain manager is now required. This manager is currently unable to spend time on the slides to review them. Although this is being actively followed up on, there is no firm time frame as to when all the compliance requirements will be satisfied. Thus, painfully, I have to temporarily withdraw the Kapital presentation.

Nevertheless, there is always a way to make things better. We have scheduled an alternate presentation. It is called "A Pattern of Perception", and it develops ideas I found at the 2006 Coding Contest much further. AGC will use the pattern of perception. AOStA is another example where it will come into play. I showed a beta of it in Buenos Aires last year and reception was very good, so I hope you will enjoy its first formal presentation in final form as well.

The official abstract is below. See you there!

A Pattern of Perception

The pattern describes a way to make explicit a model by which one can explain how the interaction between observers and their environments occurs. The pattern applies to a complex system, or information manifold, under observation and a player interacting with it. A game is an example of such an information manifold. A program playing the game is an example of a player. An adaptive compiler (player) observing the execution of a program (information manifold) to modify the performance of the program is another example.

Without an observer, the information manifold is nothing but a blob. The player, however, is able to make sense of that blob by following a sequential approach of perception, processing, and action. The processing is based on the individualization of distinctions that capture relevant pieces of information. These distinctions are evaluated and a strategy is applied so that the objectives can be met. As a result, the player reacts, influencing the manifold, and the cycle starts again.

The player is divided into an interface, eyes, hands and strategy. The eyes of the player discover some few relevant distinctions (the ball, other players, etc). Based on his strategy, the player selects some objectives to be met. Finally, his hands perform some action as throwing the ball in some direction, guided by the objectives.

Andres Valloud, until recently a consultant at JP Morgan in New York City and now lead Smalltalk virtual machine engineer at Cincom Systems, will describe the pattern of perception. He will also discuss a framework in Smalltalk that implements the pattern that has been used in games and an adaptive approach to garbage collection self-tuning.

Sunday, February 11, 2007

Kapital at San Diego State University on February 16th


The presentation on Kapital is set to happen as follows.

When: February 16th, at 1pm.
Where: San Diego State University.
More precisely: room GMCS-308.

GMCS stands for Geology, Mathematics and Computer Science.

If you go here, you can also make maps of SDSU's campus. As you can see, the GMCS building is next to parking lot A.

See you on Friday!

Thursday, February 08, 2007

Kapital is coming to a town near you!


On February 16th, I will be giving a presentation about Kapital at San Diego State University. Kapital is JP Morgan's official interest rate derivatives risk management system, written in VisualWorks and GemStone.

As James said, I was working at JP Morgan until recently. I am very happy to have the opportunity to talk about an application that I really care about. I hope you can attend and enjoy the presentation as much as I enjoyed interacting with Kapital.

Many thanks go to the whole Kapital team, in particular to Emre Oguz and Paul Bratton, for making this possible. In addition, I'd like to thank Roger Whitney for arranging this at SDSU.

The details are being finalized as I write. I will provide an update with directions and time as soon as they become available.

See you there!

Friday, February 02, 2007

Windows Vista in the subway too

I rode the PATH train to work one last time today. Between 14th and 23th street stations, the tunnel wall has a very long series of TV projectors which are synched to the train cars passing by. The result is that you get to watch TV for about 20 seconds max while on the subway.

SubMedia, the company operating the projectors, recently replaced a Honda Fit ad with a short video of Windows Vista. First you see the round Vista logo, then some windows being stacked in 3d, then being um browsed by z-order, and then the Vista logo again.

I had a really hard time thinking that this video was supposed to make you go "whoa!". Sorry guys, ain't working :(.

Thursday, February 01, 2007

Vista, SuperFetch, ReadyBoost, blablabla

So I read all these interesting (?) benchmarks about Vista's enhanced caching capabilities. The two ideas are as follows.

  1. Use some sort of LIFO queue to sort frequently accessed files, then cache them in advance hoping they will be accessed.
  2. Use the same queue, but store files in a USB drive which will on average transfer the same information quicker than the HDD.
Ok so this is a desirable feature now? This, in the context of what I also read from Toms Hardware: that just because you are logged into Vista, your box is using 1GB of RAM. So clearly, you need 2GB of RAM for Vista to execute your applications properly*.

To begin with, I do not even think that 1GB of RAM makes sense with Vista, performance wise. Take your 1GB box and then purchase a (large) USB drive so you can cache some HDD files? No way. You use the money you'd spend in the USB drive and purchase another GB of RAM instead. Little extra investment, huge difference. And as the benchmarks show, there will be no noticeable difference then.

So fine, let's say you put up with your 1GB of RAM and decide to stick a USB drive in the machine forever. First hint: the most you will save executing Toms Hardware tests is the astonishing and incredibly high amount of 2.3 seconds.

Second hint: all this is made necessary because of bloatware, and bloatware is not what we need. For example, I do not think Microsoft Word / Acrobat is able to create a PDF draft of my book starting from scratch in under three seconds. Yet that is what LaTeX can do for me.

Third hint: waiting for applications to start up represents how much percentage of your time? And if you have a laptop and you let it hibernate, how much is that already tiny percentage reduced to?

So to me, this is just more of the same marketing spin that has been applied for ages to convince consumers they need something they do not.

For example, local calls are not free in Buenos Aires. The telephone companies charge you by the minute even if you call your neighbor. Back in the days, each time unit lasted 4 minutes at night. First, the rates were hiked so the time unit lasted no more than 2 minutes, plus the rates went up per time unit. So, first things first: you get screwed.

To make up for this mess, the telcos set up special numbers so people could call their ISPs and still use the Internet. They claimed that the rate being offered for Internet purposes was much cheaper now, so this was good, so the idea was that you'd forget about the rate hike. And what was the rate for the ISP phone calls? What the rate was before the rate hike went into effect.

Thus... bigger programs that are bloated to say the least require unnecessary complexity that artificially raises the value of this tiny accomplishment that "your application starts quickly" is.

IMHO, not something to be especially proud of.

* Toms Hardware also reports that currently Vista is slower than XP to do application work, sigh. There is also the assertion that "no version of Windows has been more efficient than the previous one".