[fab] final rough draft
Max Spevack
mspevack at redhat.com
Tue Aug 15 15:46:19 UTC 2006
Looking for any last feedback, since there's no questions unanswered at
this point.
I'm going to get lunch, and then I'm going to read through for typos, and
then I'm going to hit send to /.
--Max
--
Max Spevack
+ http://fedoraproject.org/wiki/MaxSpevack
+ gpg key -- http://spevack.org/max.asc
+ fingerprint -- CD52 5E72 369B B00D 9E9A 773E 2FDB CB46 5A17 CF21
-------------- next part --------------
Hi everyone. I'm looking forward to answering all of the questions, but
before I start diving into that, I guess it would be useful to give a
little bit of perspective about me and my role within Fedora and Red Hat,
because it will offer some context around the things I have to say.
The Fedora Project, as many of you know, is a partnership between Red Hat
and the OSS community. The highest level of decision-making within Fedora
is the Fedora Project Board, a group that is empowered to make the
decisions about Fedora policy, to set priorities, and to hold the rest of
the Fedora sub-projects accountable for what they are doing. The Fedora
Board has nine members, five of whom are Red Hat employees, and four of
whom are community members. That breakdown is not set in stone --
that's just what we started with. It is my hope that down the road, the
majority of the Board will be Fedora's community leaders.
In addition, the Board has a Chairman, and that person is whoever happens
to hold the position of "Fedora Project Leader" within Red Hat -- since
February of this past year, that's been me.
As much as possible, we try to conduct our business within the confines of
the Fedora Advisory Board, which is a larger group (about 50) of the most
prominent contributors to Fedora. This is an open mailing list with
public archives and open-posting, and its participation is strong both
from @redhat.com and community contributors.
For more information about the Fedora Project Board, please <a
href="http://fedoraproject.org/wiki/Board">see our website</a>.
In the spirit of complete transparency, a word about the answers:
All of them were composed directly by me -- it's my voice and writing
style that you're reading. But, I didn't answer them all by myself
without speaking to anyone else. I discussed some of them with folks on
the Fedora Advisory Board mailing list, with various colleagues at Red
Hat, and a draft of the responses was shown to Red Hat's corporate
communications team (not because they have any editorial control over what
I say, but as a sign of respect) and a draft was also shown to Fedora
Advisory Board.
=========================
1) Why such a divide?
(Score:5, Insightful)
by dsginter
It seems to me that 'Linux should be Linux'. Rather, we're seeing articles
about one linux distro killing another. We never see "Windows Professional
is killing Windows Home". IMHO, Ubuntu's success should be a boon for all
Linux distros.
Unfortunately, package management seems to be the great divide. What are
you doing to bring One Package Manager to all Linux?
A)
I agree with your initial comment -- one of the great powers of OSS is
that when you have a strong upstream in place that is always having
changes fed back to it, success for one distribution translates to success
for all distributions.
When you look at the landscape of all the many Linux distros out there, it
isn't surprising that there's some level of competition among them. Most
people want to feel like they are the best at what they do, and a certain
amount of competition among distros is healthy. It keeps people
innovating, it keeps them working hard, etc. Personally, I think it's
important not to lose the perspective that in the end, everyone who works
on OSS -- regardless of whether they run Fedora, Ubuntu, Slackware, or any
other distro -- is ultimately working to promote more or less the same
core set of principles (more on this later).
To speak directly about Fedora:
First, we believe very strongly in working with various upstreams. The
diff between any package that we ship in Fedora and the upstream version
is as small as possible at any given time, and we are constantly
submitting our patches and changes upstream for consideration.
To your point about "one package manager to rule them all" -- well, I
think it's an admirable goal. Do I think it would be a good thing for
Linux to begin to standardize on a single package manager? Yes, I do.
Does Red Hat have strong ties to RPM? Of course. But what does that mean
for Fedora? Well, Fedora is also tied to RPM (yum is our
application-layer package management tool, with RPM providing the
lower-level work) -- but that doesn't mean that the Board is unopen to
considering the idea of change. RPM is the reality of the moment. If
there's a better solution that gains a critical mass of Fedora engineers
who are interested in experimenting with it, then we will try it out.
You ask specifically what Fedora is doing to bring about a "one package
manager to all Linux" -- well, I guess there's a couple of directions that
Fedora could go:
1) Try to convince anyone not using RPM to do so. I don't like that idea
very much -- if RPM is the tool you want to use, feel free. If you've got
something that works better for you, that's fine too.
2) Fedora could abandon RPM in favor of another package manager. Like I
said -- if Fedora engineers want to start the "Fedora
$OTHER_PACKAGE_MANAGER Project" and see how far they can get and how the
technology works, that would be a great learning experience. We're set up
in a way that a project like that could be possible, without getting in
the way of the mainline Fedora releases.
3) Try to create something entirely new, that everyone will love. Call me
cynical, but trying to build a consensus before you actually have any code
just seems like a waste of time.
I guess the "problem" with package managers is that they are so integral
to the rest of a distro that it's a major endeavor to switch them. One
reason is that a switch of that kind would break the upgrade chain.
Technical challenges like that lead to a high level of inertia, and
therefore requires a tremendous added benefit that is gained by making a
switch.
=========================
2) Drivers Vs Linux
(Score:5, Interesting)
by eldavojohn
A lot of people I talk to say they don't like Linux due to lack of driver
support. Is there anyway you see this problem being eliminated? How do you
court vendors to support their hardware on your flavor of Linux?
A)
Linux enjoys a large amount of driver support in general, but proprietary
software drivers remain a large problem. From the perspective of Fedora,
our stance is clear. It has and will always be our goal to create a
distribution that is 100% free and open source.
Fedora Core 5 and Fedora Core 6 Test 2 have tremendous amounts of driver
support, as a result of the work of many people within the community, and
also as a result of the millions of dollars that Red Hat has invested
into certification, testing, and development over the years, which has
played a role in getting things to the point at which they are right now.
The rub, of course, is that having a giant pile of open source drivers is
wonderful, but as soon as a single user runs into a single piece of
hardware that doesn't work for them under Linux, then all of a sudden
Linux "driver support" is terrible. I'm not saying that's what the poster
of the question thinks -- I'm just saying that's a general problem of
perception that Linux deals with.
One way to deal with this, or course, is to vote with your wallet. When
there are drivers and hardware that don't work under Linux because of
proprietary limitations, state those limitations honestly and
transparently, and let the users decide whether or not to buy them.
From Red Hat's perspective as a company, we do a lot of work with hardware
providers -- Dell, HP, Intel, Fujitsu, etc., and that work is often done
directly in Fedora. We've worked with Broadcom on network drivers, and
with Promise on SATA. We're hoping to bring more and more into the fold,
and we're hoping that the open testing and certification system that is
being developed (and was announced at the Red Hat Summit) will allow
partners and individuals to help us bring even greater driver support to
Fedora.
But ultimately for Fedora the goal is free software. Including support
for proprietary drivers in our distribution would violate our core
principal of freedom in software. We choose to adhere to that principal,
even when that means that technical problems remain that could otherwise
be made to disappear.
By choosing not to ship any proprietary or binary drivers, Fedora does
differ from other distributions. Ubuntu is one example, as there is very
strong language about freedom, right up until the moment at which they say
that they include binary drivers on their CDs and in their repositories.
http://www.ubuntu.com/ubuntu/philosophy
http://www.ubuntu.com/ubuntu/licensing
=========================
3) What's changed?
(Score:5, Interesting)
by KDan
You mention that opinions are rooted in the world of 5 years ago. What do
you think has changed in the linux world since then, and how does it
affect Fedora development?
A)
I remember walking into the lobby of Red Hat's headquarters two years ago
when I showed up for my job interview, and in big letters, the first think
you see is "First they ignore you, then they laugh at you, then they fight
you, then you win." Gandhi wasn't talking about Linux, but I am inspired
by that quote every day.
In the last five years, Linux has moved along down that path, most
certainly. Market share increases. Mindshare increases -- I see open
source mentioned in magazine and newspapers when it never was before.
Name recognition for Linux and why it is different (and better) is
becoming more ane more mainstream. Far from being ignored, Linux is
getting more and more support, and in turn the fight against it
intensifies as well.
In the technical realm, the changes in Linux in the last five years have
been extraordinary. I don't think I need to enumerate the differences
between a linux distro in 2000 or 2001 and today, certainly not for this
readership.
Another gigantic change over the last five years is the fact that the PC
is no longer the king of the electronic device. There was an article in
'The Economist' last week about this, discussing the fact that PDAs have
such a huge market penetration now, that many tasks that used to require
your desktop or your laptop are moving toward portable computing.
Breaking into that part of the computing world is something that is a
possibility today that wasn't really a consideration five years ago.
Fedora needs to change. It needs to be less bulky, and more customizable.
The actual "core" packages that are required to get a system up and
running should be broken out, and applications layered on top as users
need them.
In Fedora Core 6, the installer will be able to reach out to any
network-accessible repositories and pull in packages from them, which is a
step in the right direction, and also is a big step in breaking down the
distinction between Core and Extras.
=========================
4) Worst Aspect of Fedora?
(Score:5, Interesting)
by eldavojohn
On the Fedora Project website, there are plenty of reasons listed for
Fedora to be your operating system of choice. In your eyes, what is the
most lacking aspect of Fedora as it exists today?
A)
In my opinion, the most lacking aspect of Fedora as it exists today is the
separation between Fedora Core and Fedora Extras.
For those of you who aren't familiar with what that means, I shall
explain:
Fedora Core is a set of packages (right now about 2200) that is completely
self-satisfying from a dependency perspective, and is the pile of code
that we ship, for example on the Fedora Core 5 DVD, or in our bittorrent
tracker.
Fedora Extras is another set of packages (more than 3000) that is
installable adjacent to Fedora Core, but isn't shipped on the media.
Greg DeKoenigsberg worked to kickstart Fedora Extras in early 2005, and
since then it has become arguably the most successful part of the Fedora
Project. It has thrived under the leadership of Thorsten Leemhuis and the
rest of the Fedora Extras Steering Committee. The packages in Fedora
Extras are maintained by whoever is capable of stepping up and doing the
work, regardless of whether or not they are employed by Red Hat.
The reasons for the separation between Core and Extras have to do with
build systems, CVS locations, and artifacts of antiquated Red Hat
attitudes toward Fedora, as well as antiquated processes.
I would like all of that to change. I would like for the Core/Extras
distinction to go away, and instead be replaced by the idea of a Fedora
Universe, which is a giant pile of packages that are blessed by Fedora,
and any subset of those packages that produces a functioning OS can be
called Fedora.
I believe that we will get there in time, but it's not an overnight sort
of change. Right now is the time during which we should be planning how
we can achieve a goal like that, and it's my hope that as RHEL5 stabilizes
and Red Hat engineers have some cycles free up, we'll be able to get some
of the work done on the Red Hat side of the fence that is required.
Separate from that, we would love to have more contributors. People who
want to work on code (especially code that isn't package maintenance),
documentation, infrastructure, and artwork. People who are organizers and
who want to be leaders. That's not to suggest that we don't have
contributors today who have those qualities and skillsets, but there's
more than enough work to go around.
=========================
5) Vista a Problem?
(Score:5, Interesting)
by eldavojohn
Do you view Vista as a threat to your user base? Do you or people on your
team ever change your mind about things or let looming Vista influence
your decisions?
I'm hoping that Linux distros are not pressured into adding unneeded bells
and whistles in a desperate attempt to compete with Vista. Are you
invulnerable from this mentality?
A)
Truthfully, everything I know about Vista I learned in two places -- right
here on /. and on the mini-MSFT blog. I don't particularly pay any
attention to it, and I can't really tell you what is or is not supposed to
be in it at this point in time, or when it's going to ship (insert your
jokes here). I used to have a XP partition that I'd boot to for gaming --
probably not unlike a good number of /. folks -- but it's been over a year
since I blew that away and I haven't looked back.
In terms of getting people to use Linux instead of proprietary operating
systems -- I think that battle is best fought in the world of people who
are new to computers. People will tend to be loyal to the first thing
that *just works* and doesn't cause them pain. Making that first
experience for people a Linux one as opposed to a proprietary one --
that's the challenge.
By the way, I'm not suggesting that you can't show long-time proprietary
software users the light of open source, but it's a much more gradual
process: "Another Internet Explorer exploit, huh? Hey, have you heard of
Firefox and Thunderbird? Let me help you set them up, you might like it."
=========================
6) NTFS support in Fedora/RedHat.
(Score:5, Interesting)
by Anonymous Coward
If Fedora is actually not controlled by Red Hat anymore, and Fedora is
user-oriented, why are both the only general-purpose GNU/Linux
distributions that disable the NTFS driver from the Linux kernel?
Users do need this option (unlike RedHat's customers, which are
organizations as far as I know), and for evidence, Linux-NTFS is one of
the projects with the most downloads on sourceforge.
I would like to add that NTFS is part of the mainline kernel. Compiling it
as a module will cause it to not take any memory resources other than the
few kilobytes on disk that any un-used hardware module is taking, unless
of course the user has a mounted NTFS partition.
RedHat's reason for disabling NTFS support was that RedHat is a US-based
organization and that they fear patenting problems from MS. No law action
was ever taken, and no actual patent was referenced. As far as I know,
NTFS is not even patented or patentable. Fedora is not RedHat as you say,
so this old reasoning is not exactly valid for Fedora. The IBM/SCO saga
also cleared the issue about patents in the mainline kernel.
Unless Fedora will change this simple flag in the kernel config file, I
assume it is still controlled (and not only sponsered as some would say)
by RedHat.
A)
Heh, the actual question asked is a reasonable one. I think it's sad that
it has to be surrounded with such vitriol.
First of all, I am not a lawyer. The AC who posted didn't mention his
background, but I'm guessing that he/she INAL also.
FACT -- Red Hat retains legal liability for the Fedora Project. The
Fedora Project is not a separate legal entity or organization. The Fedora
Project receives a tremendous amount of resources (people, money,
infrastructure, etc.) from Red Hat.
FACT -- If you are a proprietary software company looking to exercise some
patent litigation against an open source software company, Red Hat might
not look like an awful choice.
In the past, Red Hat's counsel has been uncomfortable with enabling NTFS
support in the kernel due to the fact that the implementation is well
patented by Microsoft. Recently, the kernel has become protected by the
OIN and the question of NTFS in our kernels has been raised on the Fedora
Advisory Board recently (within the last month or two). When we have an
answer regarding that, the analysis and result will also be published
transparently for people to comment on and discuss.
=========================
7) Dependency hell
(Score:5, Interesting)
by Tet
The introduction of yum has vastly improved the user experience when
installing software, or updating existing packages. However, it's brought
with it a new kind of dependency hell. For example, if I want to install a
PostScript previewer:
% yum install evince
[...]
Installing:
evince x86_64 0.5.1-3 core 773 k
Installing for dependencies:
nautilus x86_64 2.14.1-1.fc5.1 updates-released 3.9 M
nautilus-cd-burner x86_64 2.14.2-1 updates-released 414 k
That's clearly wrong. I only want to install a PostScript previewer. Doing
so should not require a filemanager (which I don't need or want), and
certainly not a CD burner. But these are added as dependencies due to the
clumsy packaging that seems to be increasingly prevalent in Fedora.
Perhaps (and I remain unconvinced) there's some aspect of evince that can
make use of nautilus being present. But if so, I haven't seen it. I could
well believe that nautilus could make use of evince, but not really the
other way around. But assume for the sake of argument that it can use
nautilus. That still isn't a reason to have it depend on it.
Dependencies should be packages that are required in order for another to
run, not packages that will merely enable additional functionality. In
this case -- the prime function of evince is to view documents, which
isn't significantly enhanced by having a file browser present.
Fedora is still my distribution of choice, but it's becoming increasingly
hard to use for those of us that prefer to run with a minimal system due
to the way that the dependencies have been getting out of hand. Are there
any plans to fix this, or is any work already underway to do so? I
understand that some consideration has been given to providing "soft
dependencies" within RPM (like dpkg's suggested dependencies), which would
help. Is there a timeframe for this? Is anything else being done?
I quite understand the focus on getting the system to be usable for the
average unskilled user. But the impression I'm getting is that it's being
done at the expense of letting those of us that know what we're doing do
what we want. Does Fedora have a position on the type of users it's aiming
for, or is it still trying to be a general purpose OS?
A)
To your specific example, ask and ye shall see some improvement.
http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201967
To your more general question, there's a couple of things that play a
part:
Part of the dependency requirements come from the manner in which the
packages are written, in which (for example) it's far more common for
someone to install a large set of inter-related packages than just a
single package. Regardless of that, it's entirely possible (such as in
this specific evince example) that some extra work can simplify the
dependency requirements. Bugzilla is always the best way to bring issues
like that to the attention of the packagers. From there, it's just a
matter of code and time.
=========================
8) Goals
(Score:5, Insightful)
by redkazuo
While Ubuntu has a clear, selfless mission, it seems to me the Fedora
project misses this. I'm sure while Fedora was still within Red Hat, its
mission was simply commercial. "It must be good so we can make money."
That mission no longer applies, and http://fedora.redhat.com/About/
[redhat.com] almost sounds like Fedora is just a rejected part of Red Hat,
left Free so that they could attempt to profit from community
contributions.
Is there an objective in the Fedora Project? One that is clear and may
motivate developers to join? Or is it here really just to reduce costs for
the Red Hat team?
A)
Just to clarify one thing in the question first -- "while Fedora was still
within Red Hat" -- I'm not quite sure what that means, but I hope my
explanation about the Fedora Project Board at the top of my answers clears
up any questions there.
I'm really glad this question was asked, because it gives me a chance to
try to bust the NUMBER ONE MYTH about Fedora -- that Fedora is "just a
beta for RHEL" or that "Fedora only exists to make Red Hat money" or "Red
Hat doesn't care about Fedora, it's just a dumping ground for half-tested
code". I hear all of those things from time to time, and *none* of them
are true.
Let's back up for a moment -- the Red Hat Linux/Fedora Core split took
place in 2003, I believe. And while I wasn't at Red Hat during that time,
I think it's fair to state that there were some unfortunate choices made
internally about how Fedora was positioned, and because those statements
were made with a Red Hat voice, it helped to create a very strong
perception that Red Hat abandoned the community, and that Fedora wasn't
"good" for anything, or was a rejected part of Red Hat. I think there
were some people within Red Hat who were afraid that the "admission" that
Fedora was production-quality, or that Fedora was anything more than
beta-quality, would cause difficulty for the people trying to sell RHEL.
Three years later, and that perception is still very strong in certain
places -- without fail there are a few comments about that in every
Slashdot story that mentions Fedora.
And that's fine. Red Hat had a part in creating that perception, and so
Red Hat will have to work particularly hard to undo it.
The real story of Fedora, of course, is entirely opposite from the "beta
code only, not production worthy" stance.
Our mission statement is clear, and is one that I think any open-source
developer would appreciate.
Fedora is about the rapid progress of free and open source software.
That's it. We strive to produce a quality distribution of free software
that is cutting-edge, pushes the envelope of new open source technology,
and is also robust enough that it can be relied on for server or desktop
use. One of the terms that I really like, and that I think we're doing
better and better of making a reality is that of Fedora as an "open
development lab". As a user, if your priorities are cutting-edge
technology (without the nicks and cuts of a blade) and freedom, Fedora is
a great disto to use.
The second half of the story, as it relates to Red Hat's desire to make a
profit, is equally simple in my mind. Fedora is upstream of RHEL.
Fedora is also upstream of various other derivative distributions --
CentOS, for example.
(http://fedoraproject.org/wiki/DerivedDistributions)
So when someone says "Fedora is beta for RHEL" they are stating only a
very small part of what Fedora is. Fedora is the best of what works
today. RHEL is the best of what will work for the next seven years. And
the users can decide what is best for their needs.
So saying that Fedora is the beta for RHEL, and that Fedora is *only* a
beta for RHEL, is to take a purposefully narrow view on the truth.
Fedora's upstream relationship to RHEL is simply one aspect of the Fedora
Project, which stands on its own as a distribution.
http://fedoraproject.org/wiki/Objectives
In conclusion, anyone who tells you that Fedora isn't suitable for a
production server is lying. Ask wikipedia whether or not Fedora is
production-capable. And if someone tells you Fedora is "just a beta for
RHEL" or other such nonsense, and they are telling you that with
@redhat.com after their name, then they're not just lying, but they're
also doing tremendous damage to Red Hat as a company.
=========================
9) Directory Server
(Score:5, Interesting)
by IMightB
Hi, I've been using Fedora Directory Server for quite a while, and it is a
fantastic product. I read some rumours that it would be Integrated with
FC5, but sadly it was not. When can we expect this to be a standard
feature/integrated with authentication and other areas in Fedora?
A)
Regrettably, answering this question honestly also requires admission that
the integration of Fedora Directory Server and the rest of the Fedora
Project (particularly Fedora Core/Extras) hasn't happened as quickly as
many would have liked. Directory Server is a great piece of software, but
the true merging of that into Fedora Core is something that doesn't have a
lot of traction at the moment. The Directory Server community isn't
necessarily very well integrated with the rest of the Fedora community,
and therefore the two communities are in a similar state to that of the
two projects -- in theory capable of being very good together, but in
practice sort of just existing side by side, but not as closely knit as
they could be.
When will that change? As soon as we can get enough people on both sides
of that fence able to spend the cycles necessary. I can't give an exact
date, because one doesn't exist right now, so I'd rather not just make
something up.
=========================
10) Have you tried Ubuntu?
(Score:5, Interesting)
by Anonymous Coward
Have you tried Ubuntu yourself? Is there, in your opinion, something
Ubuntu does better than Fedora?
A)
Those of you hoping for some flamebait, I'm sorry to disappoint.
Yes, I have tried Ubuntu. I have played around with SuSE, though not in
any significant way for a year or so. Prior to coming to Red Hat in
August of 2004, I had always been a Slackware devotee, and my subscription
with them is still active.
So what does Ubuntu do better than Fedora?
Let me start without even mentioning the actual distributions. I think it
is clear to anyone who is looking that Ubuntu's website is in much better
shape than Fedora's. Ubuntu.com is clean, clear, and easy to navigate for
people who are browsing it, and if you dig down a little bit, you can also
get to the Ubuntu wiki, which from what I can tell, serves a similar
purpose to the fedoraproject.org wiki.
Here's the difference -- fedoraproject.org is *just* a wiki. It's got a
tremendous amount of information, and as someone who uses the site
frequently, I know how to find what I'm looking for. But it has a bit of
a learning curve before it becomes useful.
Fedora's websites are in a state of flux -- fedora.redhat.com is
deprecated, but the killing off of that site is taking longer than I would
have hoped, as there are a variety of infrastructure issues at play. Our
wiki gets the job done, but I'd like to see a more professional looking
front-end put on it, with the wiki continuing to function as it does, but
just ever-so-slightly in the background. The biggest hurdle to making
that happen -- just having enough cycles and enough people to do the job
properly.
That aside, I am impressed by Ubuntu's LiveCD, directly installable
feature. We have similar work going on within Fedora, but so far it
hasn't achieved the same level of "officialness" as the Ubuntu code. So
that's an area in which Ubuntu is ahead of Fedora.
I played around recently with Dapper Drake. Like I said, the LiveCD was
cool. The desktop -- Gnome is pretty much Gnome, Firefox is Firefox, etc.
Personally I'm a huge fan of NetworkManager, which didn't appear to be the
default in Dapper, but something like that is just a detail. I'm sure if
I were to use Dapper full time and I wanted it, I could probably get it.
This goes back to what I wrote near the beginning about the importance of
upstream. If everyone is pushing their latest work back upstream, and the
maintainers at the top level have the time and resources that they need to
keep everything in order, then most GNU/Linux distros are going to feel
pretty similar once they are installed. Which is why I think a lot of the
OSS "religious wars" don't make a lot of sense.
=========================
More information about the fedora-advisory-board
mailing list