znbailey

There was an explanation of this during the TC50 presentation and in the Q&A afterward.

From my simple understanding, this is a management application layered on top of the Mechanical Turk platform which is able to provide more robust/advanced management and analytics in terms of tracking the quality and speed of repeat turkers, as well as being able to integrate with other commoditized/outsourced labor systems. Due to being able to maintain more state concerning tasks and turkers, they are able to provide valuable intelligence about how trustworthy and accurate the results are.

It was very interesting during the Q&A when Bradley Horowitz asked the company rep if he thought that Amazon might simply do this themselves - but the presenter stated firmly that they had a pretty good personal relationship with the people at Amazon and they took pride on the turk service as really being a black box type of platform.

Regardless, the main problem this business faces (which was handily pointed out by the experts) is the issue of gaining traction - how do you make big businesses aware of this as a possible solution to their labor issues? Even at MySpace where they pay 300 people to moderate photos for obscenity they have not realized they could much more effectively and cheaply be leveraging MT for this task.

Decatur/Candler Park, GA, US

Great point. I meant to mention this in my list of cons but it slipped my mind.

The author fails to mention another very common approach that does not suffer (as badly) from the nested set approach. This strategy is called "materialized path".

In the materialized path strategy, each node/row has its full path in the tree stored in a column and that path is "materialized" or "derived" from a depth-first traversal to that node. In conjunction with the adjacency list approach this could yield much better results in most situations than a nested set model.

Taking the tree used in the article, the paths would look like (using '/' for path delimiter but you can use whatever you like):

  electronics
  electronics/televisions
  electronics/televisions/tube
  electronics/televisions/lcd
  electronics/televisions/plasma
  electronics/portable electronics
  electronics/portable electronics/mp3 players
  electronics/portable electronics/mp3 players/flash
  electronics/portable electronics/cd players
  electronics/portable electronics/2 way radios

Alternatively it would make more sense to base the materialized path off of an immutable field such as the PKey of the record (in the same order as the paths above, from the article):

  1
  1/2
  1/2/4
  1/2/5
  1/2/6
  1/3
  1/3/7
  1/3/7/10
  1/3/8
  1/3/9

This has the following advantages:

  * Node insertion/deletion overhead is low / does not require updating other nodes unlike nested set model
  * Query for subnodes is easy: "select * from tree where path like 'electronics/televisions'". To get only direct subnodes add a parent_id specifier.
  * Getting the full path for a node is easy: it's built into the node itself
  * Node depth is easy: count the number of path "segments"

Downsides:

  * Moving sub-tree requires updating all sub-node paths

Hope this helps someone!

Cheers, -Zach

It looks like they announced they will be making a concession to bring back at least some of the old behavior:

http://blog.twitter.com/2009/05/we-learned-lot.html

The jist: they're bringing back the old behavior only when the message starts with @username (and hasn't been created clicking the reply icon).

It sounds like they made a technical change which trumps the actual message content with a piece of reply-to metadata, which is either explicitly included by the client or parsed by their infrastructure.

Previously I believe it was "dumb" and didn't examine anything, and it was up to the client to filter based on the content of the message when it came to showing half-conversations.

We use it where I work and the designers like it a lot. It works very well to build cross-browser layouts in a very quick and simple manner and has drastically reduced the amount of time it takes them to cut PSDs into raw XHTML/CSS.

One downside you will quickly notice is your markup becomes much less semantic and a lot more cluttered with CSS classes. This is simply by virtue of how all "grid" CSS frameworks like this work - they set up a ton of generic styles that are impossible to be symanticized. With blueprint you start to see a lot of classes that look like "span-XX last" or "push-XX" or "pull-XX".

There are a number of ways to get around this, however. The first is a tool that comes with blueprint that lets you set up a yml configuration which maps semantic ids to these generic CSS classes. For instance you can say "#footer => span-24 last". This will then generate the CSS in a semantic manner so you can continue using id references in your XHTML.

The second way, and this is the approach that I would highly recommend, is to go ahead and use some sort of CSS metalanguage/compilation tool such as SASS (Syntactically Awesome Stylesheets - http://haml.hamptoncatlin.com/docs/rdoc/classes/Sass.html). Chris Eppstein has done a fantastic job with the Compass CSS Framework (http://wiki.github.com/chriseppstein/compass) which builds on top of SASS, which allows you to do really amazing things with CSS.

The compass framework he has built allows you to do things like define CSS constants (colors, font sizes, etc), perform unit arithmetic, as well as define abstract classes or groups of properties called "mixins" which allow you to abstract out a lot of your repetitive CSS properties. It is also a whitespace sensitive DML which makes the highly nested nature of CSS selectors a breeze to work with. It also already comes with modules for blueprint and other grid frameworks!

I would check out his talk about it here if this sounds interesting to you:

http://pivotallabs.com/talks

HTH!