Trends - how to save Fedora ?

[Thomas Cameron]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 11/13/2011 08:30 AM, Alan Cox wrote:
>> Erm, no. Each Fedora release has brought in numerous technical
>> improvements. Virtualization, clustering, directory services, more and
>> more features and performance per release.
> 
> That's a politicians answer. It's completely ignoring the point raised.

Respectfully, Alan, it's not. Fedora's charter has always been a place
for new technologies to get *tested* and considered for inclusion in the
paid distro, RHEL. Are there warts? Obviously. 0-day releases buggy? In
some feature sets - absolutely. But in general, I've found each release
of Fedora to contain more and better features than the previous. I make
no claim it comes out perfect - it certainly doesn't. But the OP's
intimation is that each release is a bigger train wreck than the last,
and I just don't see that as being the case when you take into account
how many changes and new components there are.

To be clear - I don't have the statistics and I'm not going to take the
time to go chase them down, but if you were to look at the rate of bugs
per feature set, I would be surprised if they were higher today than
they were with e.g. FC6.

> It doesn't matter how many features a new release has if it doesn't even
> run properly on lots of systems. 

Ah, but what is "lots of systems?" Back in, e.g. the FC6 times, there
were MANY fewer users, with a corresponding lower number of systems.
ISTR people were a lot more careful about buying hardware that Linux
would work on. Today, with so much larger a user population, and so many
more systems, is the percentage of problems any higher than the FC6
time? I don't know, but I'd be surprised.

> Most of the features are also
> irrelevant to most of the users. In F15 you could at least make the case
> that Gnome3 was relevant to users even if some hated it and chunks of the
> code were at best prototype state. (and I'd note the Phoronix survey data
> suggests that Gnome 3 is rather more liked than some might think from
> list traffic)

They are very relevant to me, but to be fair, I do work in a larger
enterprise computing environment. To me, Fedora is a fantastic platform
for seeing what's coming. Again, I won't say there aren't issues with
each release. I just don't think that it's really getting worse. In my
experience (I've installed Fedora on literally hundreds of systems from
Dell, HP, IBM and tons of whiteboxes, running both desktop and server
loads), I've had *significantly* less hardware problems than older
versions of Fedora.

> But clustering and directory services, like forcing LVM on hapless end
> users are really irrelevant to most. LVM wasn't a big deal for those who
> knew better - disable it on install and your disk I/O improves, and
> its become vaguely relevant with crypto. All of this is painting the
> fences and hanging bling on a core product which is getting a bit
> wobblier every release

I won't argue that LVM is not the best choice for high disk I/O
workloads, but the convenience of LVM for average users likely outweighs
the performance hit.

> It's bloated

One might argue it's got more features, and you can easily trim out the
ones you don't like.

> It picks bad user defaults

Depends on who you consider the target audience. Does it choose bad
defaults for heavy I/O server use? Probably. Does it allow you to
closely configure the settings for specialty use? Yes, and I'd argue
much better than most OSs out there.

> It ships a default desktop which burns CPU horribly

OK, you got me there, but remember back when KDE 4 shipped. Don't you
remember the howls of righteous indignation?

Now my KDE-using friends are back to touting KDE as the best thing since
sliced bread, and the friendly desktop banter is back in full swing. The
only way that happened was for KDE 4 to get released, warts and all, and
the community to beat it into submission.

>> And what? All the engineers at Red Hat develop new tech in Fedora. Where
>> do you propose those new technologies come from if Red Hat splits off?
> 
> Perhaps the Red Hat engineers could QA their new technologies a bit
> more before including them ?

They are QAing their software, by putting it out into the community for
review and comment. Beware the bleeding edge, it's sharp.

> I don't buy the "big problem" claim here. Several other releases have
> been a bit wobbly especially out of the box first release. Nor do a few
> crash reports in themselves form a statistically valid sample.

See above - I am not surprised they are "wobbly," but if you look at any
Fedora release after a few weeks, it's very solid. As expected.

> I do think that as has happened a couple of times before now it's time
> Fedora spent a release or two being more conservative on new toys and
> fixing the ones it already has.

Fair point.

>>> Linux distros:
>>> http://w3techs.com/technologies/history_details/os-linux/all/y
>>
>> Without knowing a *lot* about how this information was gathered, it's
>> meaningless.
> 
> Ah the cult of Gnome defence - insert fingers in ears and keep shouting
> loudly "We can't hear you, we can't hear you, anything we don't agree
> with is biased"

I didn't say that at all, and it's crap you claim that. All I pointed
out was it's a chart with numbers and no references to how they were
gathered or interpreted.

And I'm an XFCE, not a GNOME, user. :-p

> (to be fair I note you point to some sensible stats further down)

Thanks - that was my intent. Statistics are evil beasties, and can be
twisted to prove damn near any point you want.

>> Red Hat as a company is poised to be a billion dollar company this year
>> (FY12). The FY 2006 earnings were $278.3 million.[1] That's a 4X
>> increase in just 6 years. That's *amazing* growth.
> 
> RHEL is IMHO a good product, with well thought out services around it,
> but it's not Fedora, and I really don't want to think how 'we've
> redesigned all your init scripts and broken compatibility' would go down
> in a meeting with a major banking client. I suspect 'The door is that
> way, Sir, goodbye and tell the Oracle salesman to come in as you leave'

Again - that's why this stuff is tested where Red Hat has always said
it'd be tested - out in the community in Fedora.

>> Look at things like http://fedoraproject.org/wiki/Statistics, which
>> indicate that downloads and torrents are going up with each release, not
>> down.
> 
> Be careful that downloads are a lagging indicator of success. They go up
> after you get it right not as, and they go down after you get it wrong,
> not as...

Agreed. My main point was simply that statistics in general are to be
treated very suspiciously.

My second point was that, contrary to what the OP said, my experience
with Fedora has been generally very good and gotten better with each
release. You'll never hear me say it's perfect, but it's *so* much
better than other software I work with.

Now I'm back to cussing and screaming at the Exchange server
installation I'm working on on the other desktop. Wanna talk about CRAP
software? Don't get me started. <shudder>

Thomas
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.14 (GNU/Linux)
Comment: Using GnuPG with Red Hat - http://enigmail.mozdev.org/

iEYEARECAAYFAk7AH2oACgkQmzle50YHwaApuQCgvEACoYRQNuc+r6TDQOJ+Oj1v
1I4AoJHmkfpOsPof/G2F0AhTDA8OKahf
=30w/
-----END PGP SIGNATURE-----