On the more extreme side, there is test driven development (TDD). You write the tests first, watch them fail, and then start producing the code that will get them to pass. Having co-authored bzr-upload with the TDD-obsessed bzr developer, Vincent Ladeuil, I thought that if I was going to add a new feature, I may just as well try it (again).

It rocked.

I set up the test, my carrot, and the task went from "start poking around code" to "fix this problem". With the test written, it became very clear what parts of the code I needed to change, and how the feature had to work.

The results?  in one hour, I implemented a feature that lets you ignore specific files on upload. With tests.

To ensure we didn't loose any of the conversation, he took notes and sent them to me, so now I'm passing them on for those of you interested in contributing or just knowing what features are in the pipeline.
* Create a .bzr-upload-ignore file that ignore any action for which one the paths matches an ignore regexp. Use the working tree version by default, fallback to the versioned one otherwise

* New command: "bzr upload-files FILES" to allow uploading individual files. Upload the specified files if no uncommitted
changes exist, --force overrides the uncommitted changes check.

* New command: "bzr upload-check". Walk the remote site ensuring that every file still has the same content that the local version --restore optionally restore the remote content to the local value. Optionally for remote sites implementing ssh and providing an md5 binary, the check can be implemented by comparing the local and remote md5 avoiding the full downloads.

Edit :  High level of understanding, but a strong association with "attention", "warning", and "danger".  Might be worth modifying colour or shape to distance the icon from that interpretation.

Merge :  Reasonable understanding of "merge".  However, participants were not entirely sure if the icon referred to the state 'merged' or the branches themselves.

Remove :  Icon strongly associated with "do not enter" and "delete".  The interpretation "remove" comes only in third place.  The icon is strongly evocative and might be better used to designate a more consequential or prohibitive action.

Remote bug :  A reasonable percentage of  respondents understood the "remote bug" icon.  Many, however, did not.  It appears that the key for interpreting this icon is the representation of the bug itself.  Various potential states of a bug were suggested as interpretations.  This icon could be made more explicit.

External link :  Relatively well understood.  It is worth noting that the icon has powerful suggestion of globality and reach (associated with translation, languages, internationalization, etc).   It is a very evocative icon that could be more fully exploited perhaps in another context.

What next?  We'll attempt to create new versions of the icons, run another session of user-testing, and if understanding improves, Launchpad gets new icons  \0/

In the mean time, Charline Poirier, who is in charge of user testing in our team, has created another survey with 5 more icons to help us get more data. If everyone could give this survey another spin, and create some networking effects to help spread the survey to non-Launchpad users, it would be tremendously helpful to us. Here's the link: http://www.surveymonkey.com/s.aspx?sm=6iwthaIT4FwPCsMPa1EDEA_3d_3d

The Canonical User Experience team has set up a survey to gather information on how users see our icons, so if you have a few spare minutes (it's very quick!), please take the survey and pass it on to other people, especially if they don't use Launchpad, as they will be less biased.

Survey is available at: http://www.surveymonkey.com/s.aspx?sm=8hXmjrmFS7TmQCjh7jJB_2bQ_3d_3d

Mark Shuttleworth really deserves a lot of praise for this bold and brave move, open sourcing not only the code, but all  it's history. It's a fantastic day today.

Update: yes, fully means including soyuz and codehosting, Mark has decided to release everything. The whole history is there.

See the loggerhead page:

Note that since Launchpad has the edge/production split, so the changes in the graphs are less drastic since a set of users start interacting with new code before others.

As I was walking by, I saw a painting that really struck me. It was a terrible and dark dragon eating two men, one of them is in agony while it's face is being eaten off. Quite shocking:

After looking at it for a little while, I went closer to read the description of it, which unexpectedly shocked me ten times as much:

"This gruesome episode comes from the story of Cadmus which is told in Ovid's 'Metamorphoses' (III: 1-151). Cadmus was sent by the Delphic oracle to follow a cow and build a town where it sank from exhaustion. The cow stopped on the future site of Thebes, and Cadmus, intending to sacrifice it, sent his followers to get water from the neighbouring well of Ares. They were killed by the guardian of the well, a dragon who was the son of Ares. Cadmus then killed the dragon and on the advice of Athena sowed its teeth in the ground, from which sprang up armed men who slew each other, with the exception of five who became the ancestors of the Thebans."

This got me thinking on how much first impressions are important in the user experience, but really hit me how much more important the actual content is. We tend to relay the content creation and management to "the marketing folks", when I feel it's a crucial part that should be worked on together to balance off the amount of text, with the tone in which it's written, and to ensure that we're adding value to the users' experience.

Yes, I'm starting to see UI everywhere.