Archive for January 2004

31 January 2004

Some classic papers and lectures I’ve been reading lately:

• The Magical Number Seven, Plus or Minus Two: Some Limits on our Capacity for Processing Information, George A. Miller (1956).
• How Long Is The Coast of Britain? Statistical Self-Similarity and Fractional Dimension, Benoit Mandelbrot (1967).
• Electron Tunneling and Superconductivity, Ivar Giaever (1973).

Google can find all these.

30 January 2004

David Bradley, the person responsible for selecting the Ctrl-Alt-Delete key sequence, is retiring from IBM.

About five years ago I interviewed with Dave when I was looking for a job at IBM. Actually, looking for a job there was only half of it, though even in 1999 they were already heavy into Linux, and their research teams are generally world-class.

The other half was that I needed someone to fly me to North Carolina so I could go to Linux Expo. This was a frequent technique for getting to conferences when we were in the pre-salary stages of Ximian:

Step 1. Find a company with offices in the general region of the conference.

Step 2. Apply for a job there.

Step 3. Schedule the interview just before or after the conference.

File under Advice to a Young Entrepreneur.

Miggy.

We open sourced our build system, which has many man-years of effort in it. Build buddy is primarily a tool for building packages on multiple Linux distributions; it has built-in knowledge of the major differences between the various Linux platforms, and can build RPMs, debs, and HP-UX packages from a single XML build configuration file.

We’ll probably be putting out and open sourcing all the conf files for building Ximian Desktop as soon as we figure out how to go about doing that.

There are some interesting mockups here.

I’ve been snowboarding the last two weekends, and am going again this Saturday. It’s been an awesome way of dealing with the short days and cold darkness here. Hopefully by the end of the winter I won’t be hearing my head hit the ice like a melon every run.

28 January 2004

I moved into this fancy new apartment six months ago — a couple weeks after we sold Ximian, not coincidentally — and was immediately overwhelmed with the task of furnishing it. Especially with all the traveling I’ve had to do lately.

Buenos Aires.

So I hired a decorator to handle all the leg work. Which was great: I finally had an outlet for all my crazy homemaking ideas, like a fully enclosed circular couch and 3d wallpaper made of photos. She was arranging to have people build furniture and FedExed fabric samples to my hotel room, just like in Lost in Translation. Everything was going swimmingly, if a little slowly.

Then yesterday she quit because I am, apparently, hard to work with. This is what you get for paying in advance.

“Hard to work with.”

So I’ve been looking for furniture online tonight. This lamp is allegedly ten feet tall. And this, uh, thing looks pretty cool. Too bad it costs as much as major surgery.

Alex Graveley in Rio.

Death is very hard to deal with. I’ve had an enormous amount of trouble sleeping the last couple of months, and there’s this simmering background anxiety that my closest friends will continue dropping like flies. When Evolution 1.0 came out, I gave Ettore a guitar autographed by everyone in Radiohead. Now, it’s sitting in my apartment.

Recoleta cemetery, Buenos Aires.

And whenever someone else sleeps in my bed, if I wake up first, I’m briefly convinced that they’re dead and have to shake them awake before I’m reassured.

This is irrational, this is stupid, but there it is.

Rio.

Software is reassuring. I have spent most of the month getting my teams organized, which is really quite excellent. I am dead set on making our hackers in Bangalore integral and contributing parts of the open source communities they work with; we have had some limited success so far, and we will keep pushing. There are cultural challenges and there are organizational challenges, but we are highly motivated to succeed.

In a Rio shantytown.

The GNOME board meeting this week was, as reported by others, unusually exciting. Some new tasks include setting up a repository of contributed slides, handouts, demo files and scripts that can be used to give presentations on the desktop, creating gallery.gnome.org, a place where GNOME people can put their photos, and the fleshing out of the mythical roadmap.

(That phrase, “fleshing out,” always makes me picture skin growing on a robotic skeleton.)

I’ve volunteered to write the Collaboration section of the roadmap. This is a personal jihad of mine. The desktop needs to become a more perfect tool for collaboration. Jeff Waugh has called this a “collaboration station.”

Teams of people form, unform and reform organically. Data is created, shared, changes rapidly, and grows day-by-day. People tend to look for files and other data based on people and time: these are critical pieces of metadata. Miguel sent this to me, where is it? I was working on this last week, where is the most current version? Current tools turn people into filing clerks. Receiving an attachment, saving it to your desktop — your staging area — and then moving it into the correct folder is a common operation. But then, three days later: where did I put that? And: oh shit, she sent me a new version. Now what do I do? Keep both? How do I diff PowerPoint files?

Our traditional file management, email, publishing and office tools have not kept pace with people’s increasingly heavy use of computers to interact and work together.

We need dynamic, flexible desktop facilities that map to the way people actually work together. Information control and information sharing mechanisms need to be in the hands of users, not IT departments.

I also believe strongly that the open source desktop needs to innovate to succeed; being somewhat crappier and somewhat cheaper than Windows is not enough, despite our intrinsic advantage of freedom. Innovation is expensive, and, sadly, practical software research is primarily happening in IP-encumbered ways. Spend some time browsing research.microsoft.com. We probably cannot afford to beat Microsoft to high-quality voice recognition and natural language user interface advances.

But there are cheap and clever ways to vastly improve the experience of collaboration and information management in the desktop, and we can beat Longhorn to the punch. I’ve been telling everyone about Kubi, a clever hack that creates a dynamic collaboration space using email as a transport. This is infinitely more adoptable than Groove, and it really cuts directly to the issues: teams form organically, people work on data together.

Some of my recent efforts, like Dashboard, the GNOME bounty hunt and the Evolution/EDS split are all about getting us moving in this direction. But this is really just a beginning; just a toe in the water, and hopefully a way of helping to get people thinking along these lines.

Which they are doing anyway! Talking about innovation in an open source community can be kind of scary, because you open the door for all the lunatics with their whack-ass “visionary” ideas (anyone remember beacons?). But I’m really excited to see things like Curtis Hovey working on Medusa, Christian Hammond‘s Galago presence widget project, and so on.

What do we need? What are other people doing? Microsoft has finally published all of the slides and videos from their most recent Professional Developer’s Conference in Los Angeles, though you need Windows to view them. I highly recommend that everyone get a hold of these powerpoint files and the demos and spend some time digesting them.

Here’s some crack-for-thought:

• Universal event bus. D-BUS is fast becoming the transport of choice, and thanks to Robert Love‘s work in Project Utopia, we are getting system messages shooting down this thing. But we need the entire desktop sending out events: when the user gets new mail, when an alarm goes off, when the screensaver goes on or off, when the user’s machine goes online or offline, when a buddy goes online or offline.
• Event manager. This is a pluggable system for responding to events and turning events into persistent properties. So, when you get online/offline notifications because the user plugged in or unplugged his ethernet cable, you can toggle an ambient ‘online’ bit. Applications and users can plug in rules that are triggered by a combination of incoming events and ambient properties: if the screensaver is on and an appointment reminder fires, I am away from my desk, so send me an SMS. This is a bad example, but you get the idea. Check out Microsoft’s Information Agent for some inspiration (and some things to avoid).
• Universal metadata mechanism. When I save an attachment to my filesystem, I need to keep around information about who sent it to me. Applications and users need easy access to this information, and to query tools.
• Indexing and search. We need a flexible indexing service built into the desktop. We need automatic indexing of files, but applications also need a way to request that non-file objects get indexed.

  Microsoft’s WinFS is like a loopback-mounted filesystem built on NTFS; when an application writes to the disk, the kernel can trigger userspace code to be called. So when a Word document gets written into the WinFS, the userspace indexer gets executed and indexes it. The metadata — like author and subject — are extracted from the document and stored into WinFS’s metadata database.

  We also need ubiquitous and simple search in the desktop, across the user’s entire personal information space (a term we used a lot in the dashboard). Check out Panther: there’s a search widget in the file manager, and it searches as you type.

• Team space browser. We need a way of visualizing a team’s space. Who’s working on this “project”? Are they online or not? What documents, mail threads, appointments, bookmarks, weblogs and other resources are associated with it? I want to call an online meeting, where’s my shared whiteboard?
• Dashboard. Why can’t my computer automatically show me things that will help me with what I’m doing, instead of making me search around for them? We got here first, but Microsoft calls this “implicit query.”

By some fluke of capitalism, I now control enough resources to help make this happen. And together we have a chance to create the most powerful and smoothest collaboration experience ever, and to make it free.

And that’s what this is all about: empowering people through software.

16 January 2004

We‘re looking for a full-time OpenOffice.org hacker to be based in Boston. If you’re interested, mail Michael Meeks to apply.

14 January 2004

“I’m an old guy, and I’m rich. And there’s nothing they can do to hurt me.” — Paul O’Neill

12 January 2004

Back from Buenos Aires and Rio de Janerio. A fantastic trip.

I have a lot of catching up to do.

Some recent exciting news is that Larry Ewing is going to pick up F-Spot and start maintaining it where Ettore left off. He’ll be working full-time on it after Evolution 2.0, so we will have a world-class photo management tool in GNOME. It will also be nice to see this get integrated with Robert Love’s recent HAL/Utopia work.

Another Rhett Creighton creation: pat2pdf.com.