Speaker Mike
Time 2004-01-17 10:00
Conference LCA2004

good-morning Mike
the morning after the night before
Saturday morning, the last day

wonderful keynote speaker

hopefully I wont put you all to sleep

$2800 for charity

accommodation; hopefully have checked out

  • no - I am staying until tomorrow

Signed T-Shirt

  • $3600, thanks to auctioning skills of rusty
  • ComputerBank SA; over $5000 ????
  • plus 1 year service costs for van as promised

best of talks announced lunch-time

dunk tank lunchtime, 1pm

  • $2717.45 raised for dunking Linus
  • $5217.45 dollars total

war games today after 1:30pm

Lunch BBQ. Bar Smith Lawns 12:45 Use ice-cream tickets, 1pm
Schutzenfest in Bonython Park, 40min drive

travel to airport

conference close tonight

  • more to give away
  • more embarrassing photos
  • results for FIXIT
  • Oracle GPS device draw
  • 3 GPS units will be given away
  • IBM hack fest
  • $40,000 computer to give away
  • more surprises

The state of the Open-Source Desktop

Havo Pennington
LCA2004 2004-01-17 10:00

Who else would:

  • Unify KDE and GNOME themes?
  • Window manager, simple and easy to use? Metacity

Leader of desktop integration project

Architecting the Linux Desktop

photo of his cat (drunk?)

Linux & desktop, personal opinions

assume audience is members of the desktop development community

definition of success?

  • To provide a complete desktop solution using free software
  • => the more people using this solution, the more successful we are


  • want to have more people able to use Linux
  • if only server, windows will eventually take over server market
  • challenging, profit
  • bring more people to open source

why more people?

  • allow more people to use open source instead of just geeks
  • is the right thing to do
  • prevent Microsoft dictating standard
  • allow allocation of development resources

desktop now

  • desktop happening this year, not when I thought it would
  • commercial investment into desktop finally happening now
  • strong customer interest
  • Brazil, Spain, China, numerous universities, and many others
  • longhorn being delayed => nice window of opportunity
  • longhorn == next version of ms-windows???

when are we ready for desktop?

  • now
  • Google statistics, 2% Linux users
  • OS10 defectors
  • in 3 years, even more ready for desktop
  • more and more appeal for desktop
  • always will be competition for desktop
  • may completely crush Microsoft, but may take a few years
  • next step to grow just a little bit is ???

how long will it take?

  • how we enable the next class of users to use Linux
  • when do ISVs and ISPs have to care about Linux?
  • when do we overtake Apple?
  • need to be better and compete
  • right expectations concerning timing
  • Linux development going on a long time

have we succeeded?

  • when people have to care about Linux desktop, and not just Windows desktop
  • options, competition
  • software development never ends
  • Linux desktop itself

threats from Microsoft

  • XML, longhorn
    • Windows application is a web applications
    • replacement for HTML entirely Microsoft specific
    • merge web applications and desktop applications
    • seductive as possible
  • NET vs Java
  • Patents
    • potentially dangerous to us
    • allow Microsoft monopoly
  • lock in strategies
    • File formats
    • exchange active directory
    • protocols
    • media format
    • Microsoft office file formats

challenges and prerequisites

  • what challenges will will face
  • where can be be better and compete, opportunities
  • graphic subsystem
  • X-Windows system is right basis, extensible, make it snazzy and cool
  • eye candy?
  • important, credible, needs to look better then TWM, looks do matter
  • audience “TWM rocks! OLWM rocks! Openlook for ever”
  • TWM emulation mode, no time
  • usability, smooth transition for things moving, not jumping around screen
  • graphics issues
  • technical issues, not worried about that, will take time to fix
  • organisational issues,
  • sandwiched between kernel developers and desktop developers
  • political issues are the hardest issues to solve
  • drivers need to be written
  • X development, XFreeDesktop new line of development

robustness and performance

  • something gets confused? go to command line and fix it?
  • Wrong! Never require user to go command line or syslog or stderr in order to fix problems
  • This extends to kernel, printk needs to be redirected to dialog box
  • Alternatively software works around or fixed the problem transparently
  • hard problem, spans multiple open source projects
  • get people who don’t know each other to talk to each other
  • as community, need to ensure that we mix
  • performance metrics
  • Not: this feels faster/slow
  • real metrics for specific things, make sure they don’t deteriorate
  • already done for kernel
  • avoid regressions

huge barrier

  • migration path from other platforms
  • can’t convert every user to Linux overnight
  • Linux should work with Microsoft servers and clients
  • WINE, VMWare, Citrix, open source alternatives?
  • more solutions to run windows on Linux are proprietary
  • lean away from WINE for fear it might not quite work right

hardware support

  • people building devices need to coordinate building drivers before releasing product
  • need drivers –> need market share to get drivers –> need drivers –> …
  • hardware abstraction layer

ISV and OEM support

  • cautious about jumping


  • open source community hasn’t considered enough
  • legal issues
  • install default, can’t read multimedia formats without proprietary software like Real Audio
  • open format, e.g. OGG
  • windows media


  • usability
  • usability studies
  • great usability
  • expectation, a little bit scary?
  • corner cases
  • consistent usability, for all different things you might want to do
  • usability related discussion lists, everybody has their own opinion
  • some people are more right then others
  • shouldn’t be a free-for-all
  • areas of expertise over then just writing software
  • remember: not writing software for ourselves
  • can’t assume user has same priorities that we do
  • reasonably programming languages
    • C/C++ require you do too much work ourselves
    • another language, get things more quickly
    • pointers, allocating memory, difficult issues
    • C++ more complicated but at first glance looks simpler
  • mono project
    • everything automatically language bound
  • open source Java project Red Hat
  • open source scripting languages
  • increase priority of making languages work well

Documentation of Platform

  • GTK vs QT
  • freedesktop.org, use both
  • developer question rather then impact end user
  • daunting
  • keep things locked down and unchanging,
    • not ready for lock down yet
    • alternative is to get locked down into something that is no good
  • application –> application development platform
    • platform is a way to write applications

Desktop not just GUI, whole network

  • more meetings
  • whole network infrastructure needs to be implemented
  • OpenLDAP
  • lots of schemas and information need to be stored, out of box
  • commercial support
  • Linux already accepted on server, file server, etc
  • proprietary products
  • open source solutions not as good, at least not yet

Areas that we have to address

  • 2% market share according to Google
  • opportunities to be the best

Open source

  • first reason we are different: open source software
  • Microsoft can’t buy strategy, can’t kill it
  • Microsoft can’t do open source, business model not designed around it
  • need to remain open source, or we become another Microsoft without as much money
  • no vendor lock in
  • more people, more advocates, critical to our success


  • some regions to small, and don’t get noticed by others
  • Pango support all sort of languages
  • technical problems
  • local support provided
  • exceptional difficult for anyone else to replicate by centralised development company
  • address own problems with open source
  • more people can get involved in open source
  • people solving own problems


  • government regulations require it
  • US goes further, says you have to buy best accessibility support available

consistent single platform

  • same API on laptop, desktop server, PowerPC, Intel,
  • only need to port application once
  • only need to train once
  • choice of vendors
  • hardware technologies


  • isn’t the only thing that makes open software better
  • lots of cost benefits that aren’t just saving on license costs
  • don’t have to buy server from same vendor and clients
  • choice of hardware and CPU vendors
  • choose different pieces of software, open standards

open data formats

  • Microsoft could offer in theory, but probably won’t
  • can write script to convert one format to another
  • will get worse with encryption, compression, etc

data access methods

  • broken standards
  • winfs - Microsoft
    • not a filesystem
    • data access method
    • search based on content in file
    • uniform way of accessing different types of files
    • address database, email, music,
    • uniformed API
    • synchronisation of data
  • storage project
  • Reiserfs
  • gnome projects
  • network protocol
  • can we standardise open source solution
  • open standard required


  • different between bullet point
  • real standard, not token standard
  • no vendor agenda to worry about
  • interests of user rather then interest of company/vendor
  • check for new mail in less then 5 minutes
  • shouldn’t be afraid to optimise things


  • huge advantage
  • Microsoft has developer tools
  • open source has advantages
  • customise software, don’t start from scratch
  • if you want things to move in a certain direction, move them
  • level playing field, open source available for free
  • XML, database, spell checkers, everything you can think of
  • unique advantages
  • more people working on software


  • many eyes principle
  • more eyes, more likely bugs will be found
  • need to audit software
  • integrate into process
  • limited impact of security issues
  • SE-Linux limits process, and constrains what they can do
  • how do we use this for desktop?
  • any process with access to X server, can do anything
  • Sendmail vs postfix, can we do the same for the desktop
  • encrypted filesystems
  • focus on security would be very smart


  • can be better then windows
  • run windows in lots of ways
  • NFS, IMAP, X Windows
  • fine control what users can/can’t do
  • command line
  • a lot of potential, we can be better
  • we can gain, don’t be just as good as competition, be better
  • reading email


  • integration of applications

huge laundry lists
where we stand
what issues we need to focus on
more users, more developers

port to windows, move developers
pick up developers same rate pick up market share
commercial development, positive and negative impact

  • some communities have more experience then others
  • cool having new developers
  • more even faster

longhorn will arrive soon

  • 3 gnome releases
  • Linux 2.8 out
  • OpenOffice 2.2? Probably not.
  • Shouldn’t just clone XP, we will fall behind
  • focus on strengths, not Microsoft’s strengths
  • need to be better, not “as good as”.
  • 20% people still using Windows 98
  • alternative upgrade process

open source is the only competition for Windows

  • need a bigger enough edge that Microsoft can’t copy
  • choice, no lock-in, etc
  • half open source, half proprietary == proprietary

aim high

  • not guaranteed to work
  • years of effort, has been many years, will be many years
  • if we try, we will succeed, otherwise we wont
  • direction of technology
  • strategy
  • try to make desktop succeed
  • don’t consider yourself as bystander

All right, its 11:30