Lost my voice last night talking to people about our GNOME Online Desktop keynote.
I’ll try to remember some interesting things people brought up.
Some commenters thought the talk wasn’t alarmist enough and
should have more strongly stressed the urgency of the situation. I
don’t think we need to panic but I do think a lot of hard work is in
order, in either the online desktop direction (or some other focused
direction others may propose).
Aren’t many web services proprietary? was a good question
raised during the talk Q&A. My short answer to this is yes, but
ignoring the real user benefits of these services will result in
everyone using them anyway, while not using GNOME or any other open
source software. We have to engage with where the world is going and
what users want to have. That will put us in a position to affect
openness. Taking a “stop progress!” attitude won’t help.
Even among GNOME developers, almost everyone uses
Google or Flickr or something. Expecting a wider
non-geek audience to forego these services on ideological grounds
while we aren’t even doing it ourselves doesn’t seem very reasonable
Also, web services may well not be proprietary in the sense of the
Open Source Definition, but are proprietary in effect. See
below, on the need for an Open Service Definition.
I don’t want to put my data on someone else’s server and other
security issues are a common concern. Let’s be clear that of course it
will always be possible to keep your data locally, or run your own
But I think the privacy issues are very solveable, even for
people who care deeply about them. An Open Service Definition or the
like might address these. And of course you can do strong cryptography
– though for the average consumer, the prospect of losing 5 years of
data when they lose their private key is not acceptable, Mozy is an example of a service that gives
you an option to strongly encrypt with your own private key. They
don’t default to that choice since it’s too risky for an
As with the issue of proprietary web services, though, a “stop
progress!” attitude won’t put us in a position to affect security or
privacy. If we want to affect these things, we first have to offer the
user benefits and be a project people really care about. And then we
can affect what other participants in the industry do.
Several people suggested the argument against security concerns is
“do you use online banking?” and that seems like a good point, since
most people do use it.
We need a Free Services License, Open Service Definition, Free
Terms of Service, or whatever we want to call it. I see more and
more people talking about this, even aside from the GNOME Online
Desktop conversation. Topics to cover in an Open Service Definition
might include ability to export your personal data, your right to own
your data’s copyright, etc. There may also be a requirement to use an
Affero GPL type of license. This is very open-ended and unclear at the
To me the reason open source works is that multiple parties with
competing interests can collaborate on the software. What would make
multiple parties interested in collaborating on a service? Probably a
fairly radical-sounding set of requirements. But the GPL was pretty
radical-sounding too, many years ago.
Running servers that require real bandwidth, hardware, and
administration will be hard for the open source community. This is
absolutely true. On the other hand, I can imagine a lot of ways we can
approach this, and we don’t need very much in the way
of servers to get started. As I said in the talk, if we produce
something compelling people will be excited about it and we’ll have a
number of opportunities to work with for-profit and nonprofit funding
sources to get the server problem solved. If we don’t produce
something compelling, then there won’t be a scalability issue.
There are some precedents, the main one being Wikipedia, but
I’m also thinking of the Internet Archive, iBiblio, and ourmedia.org
as examples of nonprofit services.
What about just using a WebDAV home directory or syncing files
around? If you start to prototype this, I would bet it produces a
distinctly different and probably worse user experience than building
around domain-specific services like calendar, photos, etc., for a
variety of reasons. But it could be part of the answer and is
certainly worth prototyping.
(This post was originally found at http://log.ometer.com/2007-07.html#18.2)