Cafe Retornym

Is "drip coffee" a retronym?

Outsourcing 2.0

CodyB over at Mission Mountain Tech has a project going which uses Amazon's Mechanical Turk to do feature extraction from geospatial imagery, and as a Friend of Cody I've had the good fortune to be able to present a talk on this project a few times now (GeoWeb, CUGOS, etc). [slides now published]

Satish from ESRI termed this "outsourcing for the Web 2.0 generation" and I've been using that as a tagline for the idea of using (some might say exploiting) human capital for projects that any right-thinking engineer would immediately assume should be automated. In Cody's case, finding swimming pools in Dallas.

This morning I stumbled across an article on the Pandora "radio station" which pays (living, breathing, human) musicians to process their music selections -- as opposed to, say, last.fm which uses a social network graph.

I'm now on the hunt for other cutting edge projects and services that have humans in the loop.

WPF and C# notes

I've been learning C# and the .NET framework and WPF in the background lately -- most recently reimplementing "jiffy", my JPEG 2000 dump utility.

Some random notes as I climb the learning curve:

• Petzold's book really disappointed me. These two quotes from an Amazon reviewer sum it up nicely: "I prefer sample applications to be shorter in order to illustrate the point. I found myself wading through code listings that had very little to do with what I thought the lesson was about" and "I felt that the explanations spent more time on how the sample application worked as opposed to going over the WPF technology that was supposed to be covered by the sample". I'm not looking for a dumbed-down version, no, but some screen shots or diagrams, please! Anyway, Sells' book is the one I've been toting around with me.
  
• Liberty's C# book is a good intro (I've rarely gone wrong with an O'Reilly title), but doesn't go deep enough for an ex-compiler wonk. I need Books Two and Three, probably a C# language reference and a Framework reference.
• A small thank you to John Skeet and his utility classes -- I'm using the EndianBinaryReader. (suggestion: namespace his classes a bit more, e.g. from MiscUtil.* to Skeet.MiscUtil.*).
  
• Thanks also go to John Stewien for his multi-select tree control. (That behaviour really should be part of the framework itself.)
  
• ObservableCollection seems to live in WindowsBase.dll. At least I wasn't the only one stuck on that.
• Cold-start app startup times are waayy painful. I've seen a few blog posts on this topic, I try some of the ideas myself at some point.
• Building produces lines in the output pane of the form: "C:\WINDOWS\...\Csc.exe /noconfig /nowarn:1701,1702 /errorreport:prompt ..." Alas, using F4 to cycle through the compilation errors line by line always "hits" this line. Why is that? Can I omit that line somehow as a build/config option? (F4 is able to skip other kinds of lines, like "Compile complete -- 2 errors ...")
• Refactoring is nice. More kinds, please!
• "AddReference..." takes a long time to open the dialog. Why?

GeoWeb 2008

Abbreviated trip report over on the LTI blog.

Gartner's Seven Grand Challenges

In this month's CACM, there's a short piece highlighting the "seven IT grand challenges that, if met, will have a profound economic, scientific, and societal impacts" according to Gartner.

1. manual recharge of wireless devices
   
2. parallel programming
3. non-tactile, natural computing interfaces
4. speech translation
5. reliable long-term digital storage
6. increase programmer productivity by 100-fold [CAMC got this wrong, they said "100%"]
   
7. identify the financial consequences of IT investments

My knee-jerk reactions are:

• 2 and 4 definitely.
• 1 and 3 and 6 maybe.
• 5 is a good problem to solve, yeah, but I'm not sure if it meets what I think of as Grand.
  
• 7 is probably not on the right list. How is this different from, say, understanding where my marketing dollars go?

Rebirth of CACM

CACM is apparently trying to reincarnate itself, and so they sent me a free copy this week; this is probably related to the demise of Queue. CACM is trying to appeal to a "diverse mix of researchers, practitioners, and educators" -- well, okay, so that matches my profile at about (0.20, 1.0, 0.20), given my nonpracticing interest in research and CS higher education.

And I must admit, the articles I skimmed were really pretty good. I was a little dismayed to see Yet Another Interview With Knuth, but it was actually a nice behind-the-scenes approach. And the ISCA review made me more than a bit nostaligc.

But for $100/year membership? Probably not, I've no real need to join ACM, do I? If they had a magazine-only rate, that might be appealing.