Rethinking the Linux Distribution
Subject:   Author's Analysis of Results
Date:   2007-07-03 04:39:12
From:   georgebelotsky
Almost two months have passed since the publication of the article.
As with any experiment, it is necessary to analyze the results. Now
is a good time to do so. The initial wave of comments has gone down,
so it is possible to review, and draw conclusions. In addition to the
remarks posted to this forum, I have searched for posts on other
sites, and incorporated them into the following discussion.
Unfortunately, it is not practical to find every single remark that
exists across the global Internet.

Discussion Highlights

First, here are a few highlights of the ideas posted to this forum.

The minimal live CD by Matthew Nourse (matt_n) would make an
excellent companion to SaaS-based applications. The CD can be used
quite safely from almost any computer -- including one with viruses on
the hard drive. The SaaS/minimal live CD is also far more secure than
storing large amounts of data on a laptop, which can be stolen. You
can find this application at

I also had a long discussion with Kuros, who made a number of
insigtful comments.

Finally, simon_hibbs proposed a fascinating roadmap, based on
virtual machines.

In addition, Bryan Dumm emailed me some information on his team's
projects related to the future of the Linux distribution. You can
download the source code from

Although the Web site is not yet finished, there are lots of
interesting ideas. You can also read their articles at

The VNC Desktop

The VNC-based desktop (using a Java applet inside Firefox) has been
the target of criticism, which does have a valid point. For purely
local applications, this approach -- as it currently stands -- is slow
and cumbersome. While I did find the system quite usable, it is not a
production-level solution for now. As the article states, the Firfox
desktop is perhaps not quite ready today.

Of course, the specific demonstration given in the article is
experimental, as evidenced by the supplied source code.
Unfortunately, this may not have been clear enough for some readers.

The idea itself, however, is not as crazy as it looks at first glance.
After all, X windows itself -- originally designed as a client-server
architecture -- works well with both the client and the server on the
same machine. This is the most common way to use X windows today, so
modern implementations are sufficiently optimized that the performance
penalty is not a problem.

Provided that the user experience is good enough for remote
applications, using the same technology for local access is appealing
-- especially if the remote applications will dominate. So, if heavy
use of such applications is the future, VNC deserves some
consideration. The technology is mature and cross-platform; at the
very least, it could serve as a testbed, or a temporary solution,
until the Web OS/SaaS variants are fully ready. Whether this is
indeed the future, or desireable, is a different topic (see below).

In a serious implementation, a combination of window manager features
built into Firefox itself, plus technology such as FreeNX (both
discussed in the article) would likely replace VNC in the above
scenario. Nevertheless it is probably too early to abandon VNC

Is SaaS a Good Idea?

There were a number of remarks that SaaS/Web 2.0/Web OS is merely a
marketing phenomenon, created just so large companies can generate
perpetual revenue via subscriptions. Some of the comments went on to
say that a FOSS Web OS platform would actually harm the FOSS

The criticism raises important points. In particular, there is a
danger in giving up control of your applications (and especially your
data) to a third party, which has the power to change the software or
its terms of use at any time.

SaaS and the Web OS, however, will be with us in some form. Today,
the Venture Capitalists and large companies are funding Web 2.0. The
massive infusion of capital attracts plenty of top engineering talent.
After all, few active engineers are sufficiently wealthy to sustain
major projects on their own. This is true even for FOSS work, which
almost always requires backing at some point.

Such dynamics are powerful enough to overcome the high failure rate of
startups, and the inefficiency of many large corporatex IT projects.
The phenomenon was clearly visible in the dotcom era. Despite the
crash, Web 1.0 was not, in the end, a failure. Companies such as
Amazon, Ebay, and even Google emerged from that period.

Thus, declining to build a Free Web OS platform will not prevent
proprietary Web OS offerings from emerging. Without the FOSS
alternative, however, there will be far fewer choices in the market.
The mere presense of a FOSS platform will tend to counteract the
natural tendency of vendors to lock in their customers. Lock-in is
especially dangerous in a SaaS scenario, so the availability of FOSS
solutions (which any ISP or small company can host) is particularly

In addition, it is necessary to keep the needs of all users in mind
when discussing SaaS. Technology enthusiasts can maintain their local
desktops by themselves, but others require frequent, ongoing support.
Paying a subscription fee may well be more reliable, and less costly,
than emergency visits by technicians, or running to the store to have
the PC "repaired". IT departments inside organizations face an
analogous situation.

Of course, it would be fascinating to discuss if some other direction
would be preferrable to the current Web 2.0 trend. Unfortunately, all
the criticism of Web 2.0 that I came across argued for the
preservation of the status quo. This, of course, is not possible.
Only dead things stop evolving. The future Linux distribution
will be different from today's. Discussing that future is
vital for the community -- otherwise outside forces will fully
determine our next platform.

Harnessing the Wisdom of Crowds

The main goal of the article was to start a discussion on the future
of the Linux distribution. To help this discussion, the text contains
a large number of references, a strong encouragement to participate,
and a governance framework to guide the conversation.

With a few minor exceptions, the above approach had the intended
effect in this forum. Nevertheless, lots of comments across the
Internet ignored the governance framework altogether. Indeed,
hundreds of posts clearly originated from people that have not even
read the article. Unfortunately, such situations are common in online

These results suggest that the ubiquitous forum may not be the best
way to harness the "Wisdom of Crowds". One of the ideas covered in
the article, however, offers a compelling alternative.

The Aviation Safety Reporting System (ASRS) has been preventing
aviation accidents since its launch in 1975. As mentioned in the
article, adoption of a similar facility in hospitals reduced medical
errors by up to 90%.

Rex Hardy, one of the architects of the ASRS, writes the following in
his book, CALLBACK


    "This book is about flying and flying people, but it is not, in
    general, filled with excitement and tales of high adventure. ...
    instead, it chronicles the problems encountered by many
    all-to-human anonymous airmen."

    Thus, you get insights from anonymous reports, which anyone
    can file. Clearly, the ASRS has made effective use of the Wisdom of
    Crowds -- starting fifteen years before the Web even existed. This is
    the key idea behind many Web services today. Can any modern social
    network or Web 2.0 site, however, match the ASRS approach -- which has
    proven capable of reducing errors by 90%?

    To help improve the Linux distribution, it would be highly worthwhile
    to set up an ASRS-style reporting facility. Unlike a bug tracking
    tool, this facility -- just like the ASRS itself -- would focus on
    human factors. What combination of features, packages, configuration
    options and documentation cause confusion or errors? Which
    distributions are most affected? How serious are the resulting
    mistakes? Do the problems affect developers, power users or
    beginners? An ASRS-style tool can answer such questions.

    This reporting facility would augment mailing lists, web forums, bug
    tracking tools and other technology that enable the FOSS community to
    function. For ease of classification and retrieval of information,
    the system would allow users to attach tags to each report -- just
    like a social bookmarking site. The system should generate a large
    amount of useful data on what are greatest problems that users and
    developers acutally face. This data could then drive automated
    alerts, provide the basis for montly bulletins, and support various
    research projects. Indeed, the original ASRS itself continues to be
    the foundation for such activities.

    The ASRS has proven itself capable of generating enormous
    improvements. The value of such a system to the future of the Linux
    distribution, and FOSS in general, should not be overlooked.

    A Note of Thanks

    I would like to close with my sincere thanks to everyone who took the
    time to read the article, and post thoughtful responses. Also, thanks
    to the Pardus team for such a fascinating Linux distribution design,
    as well as the insigtful papers on the application of Python.
    Likewise, the Mozilla developers deserve lots of praise for their
    work, as do other FOSS innovators, too many to mention here.

    Real innovation requires not only technical expertise, but also hard
    work, and quite a bit of personal courage. All of us who use FOSS
    should be grateful to those who apply their abilities so that the
    whole community can benefit. Thank You.