Stuart Herbert | Gentoo

The announcement for the Gentoo Seeds project made top article on this week’s LWN’s distro page. Go check it out

They also include a link to their article on Joe Barr’s recent article. Their article is brief, but I’d urge every Gentoo developer to make the time to read through the comments attached. It’s always useful to keep in mind how folks feel about what we do; it’s too easy for us to get wrapped up in our own little eco-system and forget the users themselves.

That’s why it’s also worth reading Slashdot’s more troll-like take on the Seeds project. There’s a couple of clear trends running through the comments attached to that posting. There’s also a cowardly attack on Seemant by someone claiming to be a Gentoo developer. If that really is from a Gentoo developer, well, all I can say is that just hope it’s not from anyone I call a friend, because I think that’s disgusting.

A couple of folk have asked me where I get my ideas from. Well, there’s this mail order company …

Okay, so maybe I shouldn’t be giving up the day job for a spot down the local Glee Club. But, seriously, if anyone’s short of inspiration, all you need to do is find a bunch of folk who aren’t part of the Gentoo dev community, and listen to their experiences with Gentoo. And listen to their experiences of having to deal with the bunch of us. We’re all too close to Gentoo to be able to see everything clearly, and as a group we’re really bad at looking inwards far too much.

Take this chap as an example, and see what your next idea to improve Gentoo might be. Whilst you’re doing so, take a good look in the mirror at your own reaction to his comments. You’ll see many of the core problems of Gentoo right there and then.

We’ve taken the idea of creating stage4 tarballs, and planted that seed as its very own project!

Gentoo Seeds is a new project, which will work on creating rich Gentoo installs for common scenarios. We’re starting off with a Gentoo LAMP Server, and are already talking about following that up with a Gentoo LAMP Developer Desktop.

We’re at the “let’s figure out how to do this” stage. If you want to join in, please do! We’re hanging out in #gentoo-php for the moment.

And we’d love to work with other Gentoo devs who have ideas for seeds of their own.

My progress towards an AMI for Amazon’s EC2 has hit an unexpected snag. This evening has been one ‘gcc internal error’ after another. Not in the dom0 machine, but in the domU machines. I wasn’t having this problem last week.

I won’t be able to order any replacement RAM until mid-October, so here’s hoping that it’s a software problem.

One of the pieces of feedback I’ve had about the stage4 releases idea is that we’d be putting Gentoo into direct competition with rPath, the “Linux appliance” company founded by ex-RedHat folks (including Erik Troan, the author of RPM).

Rather than post yet-another-long-article about this, I’m going to keep this one short (well, short for me ). I’ve got a challenge for anyone who’s interested in seeing us produce stage4 tarballs for our users. I’ve spent the day taking this challenge, and I feel I’ve learned from it.

The challenge is simple: head on over to their website, read their docs, look at the rLinux package repository, download one of their applications and play with it.

See what your answers are to these questions:

1. What is rPath doing right that we’re doing wrong?
2. In what way are rPath’s appliances better than the equivalent manual Gentoo install?
3. Would you use an rPath appliance for what you’re passionate about using Gentoo for?

Then compare them to your answers to these questions:

1. What is Gentoo doing wrong that makes you look at alternative distros?
2. To get your system right, what do you have to do that Gentoo doesn’t do for you?
3. How good is Gentoo at what you’re passionate about?

That’ll give you a personalised idea about how much of a competitor rPath is to Gentoo, and how well Gentoo really matches up to what’s important to you.

Tip to save you a bit of time: use vmware server to test their appliances. qemu doesn’t build on gcc-4.1 systems, and none of the appliances I looked at ran under Xen.

I’ve just heard that Rob Levin (aka lilo), the man behind the Freenode IRC network, died today after being hit by a car on 12th September.

I met Rob through Reuben (who worked for Rob as part of Freenode, and for me as part of the Gentoo UK conferences). Somehow he seemed to make time for everyone. Rob was always a great supporter of Gentoo, and sometimes he’d message folks and ask how things were going. Rob always gave me great advice about how to get the best from Reuben, and without that we probably wouldn’t have had the 2004 and 2005 Gentoo UK conferences.

Freenode has been our home for many years now, even moreso than the Gentoo mailing lists. Many Gentoo projects - including all the ones I’m involved in - do the vast majority of their collaboration through Freenode. Most of the folks I’ve recruited for Gentoo have come through meeting them on Freenode. Through the PDPC, Rob’s work has made a difference for us, and ultimately for all the folks who enjoy running Gentoo on their boxes.

R.I.P. Rob.

(This is my promised follow-up to Donnie’s post about one way we could bring some focus to Gentoo)

The Jumble Sale That Is Gentoo

Folks over in the Debian camp are afraid that they’re turning into Gentoo. They may not say so by using our name, but the fear is very real. Joey Hess is one such person, who stuck his head over the parapet after DebConf with his now famous quote “that Ubuntu is reducing Debian into a supermarket of components.”

Isn’t that exactly what Gentoo is - a supermarket of components?

Take a webserver as a good example - one with the company website on it. A stage3 tarball and a bit of work brings you a bootable Gentoo box. An emerge command or two and a bit of work brings you a Gentoo box with Apache/insert-your-favourite-http-daemon-here on it. The webapps overlay and a bit of work brings you a Gentoo box w/ your favourite open-source CMS on it, ready to start creating your website.

Most folks call that a webserver, and they’re right in a certain sense: it is a box that can serve up web pages to visitors. But it’s a long way short of a complete webserver that can safely and reliably sit on the Internet all day every day. The problem is that most folks stop round about the time that they get the CMS onto the box. They either don’t see that extra yard they need to travel, or they don’t want to yet another bit of work to put the bits that are missing.

We expect (and rightly so) that folks should know what they’re doing, and that they should choose the components they need from Gentoo’s supermarket of packages to build the system that’s exactly right for them. It’s the one freedom that Gentoo provides that all the other top-table distros - RedHat, SuSE, Debian, Ubuntu and Mandrake - will always fall short of.

But we could do more. If we did it right, doing more would be good for our users, and it would be good for us (as in Gentoo’s development team) too.

Stop Looking Down At Our Feet

For a start, we could give folks a leg up, and provide a stage4 tarball with all the basics done. Something containing Apache, PHP, and MySQL. A Gentoo LAMP Server tarball, for lack of a better name. Then we could go a bit further, and add in basic security measures to help with firewalling, intrusion detection, attempts to crack SSH passwords remotely, and so on. Then we could go further still, and incorporate automatic server performance monitoring, and also log file analytics. Those are big things, but we can go even further, and add in the little things like log file rotation and compression and automatic alerting.

Now we’re off to the races. Our LAMP Server tarball is still very generic - we haven’t included any software like Drupal to create web pages - but now folks can download it, slap it on a box or a VM, and all they need to do is add in the packages they need to power their website. We’ve taken care of all the platform things. That gives us a horizontal market. And we need to get there first; it’s the first step towards creating a sense of vertigo.

With our LAMP Server - which is nothing more than a tailored release of Gentoo Linux - we have something that can be customised further, to suit the needs of specific types of business. For example, take a LAMP Server, add in SugarCRM and Drupal, implement single sign-on between the two, and you have a LAMP Intranet Server that will help telesales and distribution-based businesses. Suddenly, Gentoo-powered solutions are being targetted at vertical markets, not just the horizontal ones. Now, we’re a long way away from a supermarket of components, but at the same time, we’re still exactly a supermarket of components.

Gentoo doesn’t have to go the whole hog and aim to break into vertical markets. We can leave that to entrepreneurial folks who are better at that sort of thing than we are. But we need to give them (and us) a leg up, by creating the platforms for them to build on.

In the process of creating, and maintaining a LAMP Server tarball, we’ve given the Gentoo developers involved somewhere new to look. A LAMP Server tarball isn’t just a platform for folks to run webservers from - it’s a platform for developers to learn new skills, and to see the world in a different way. The focus shifts. Not only do developers concentrate on creating great packages - on putting components on the shelf of the Gentoo supermarket, if you will - but now they’re focused on putting something into the hands of people and saying “This solves your problems.” Over here in the UK, that’s a skill called “commercial awareness”. Very few technical folks bother to acquire it; those who do are valued diamonds indeed. We add another educational aspect to Gentoo, and that can’t be a bad thing.

We Can All Do This If We Want To

I’ve picked a LAMP Server for the example because web stuff is what I do. But there’s no reason why we can’t produce stage4 tarballs like this for other groups of Gentoo users. If there’s a consistent group of users that we can tailor a Gentoo release for, and if there’s a group of developers who can sustainably create releases for that group, we have ourselves another vertical market. (The key here is sustainable - we only embarrass ourselves if we fall short of that measure).

Gentoo’s still Gentoo: it’s still source-based; it’s still based around the Portage tree; it’s still supremely flexible; and it still reflects the interests of the volunteers who kindly donate their time to it. There’s still no uber-dictator calling the shots; we still have teams pulling Gentoo in many different directions all at once.

But it also gains a visible cutting edge that it didn’t have before, because we’ve sharpened our skills and focused them on something that’s outside Gentoo - how we can make Gentoo more immediately relevent to our users. If we can learn to listen well, the direction that folks feel is missing will come from delivering stage4 tarballs to meet users’ needs.

A Footnote: Swilling Out The Sty

My interest in doing down this particular approach is firmly about getting Gentoo into more people’s lives, by making Gentoo a more relevant choice for people. But it’ll also bring another benefit too - it’ll help solve the other complaint from Donnie’s original thread - those Gentoo developers who drag the rest of us down.

Shuttleworth isn’t the only reason that Ubuntu is a success. He’s brought stability and a handy pile of cash to the table, but it’s still the package maintainers that do the work to create Ubuntu. What he’s done is create a group that attracts a certain calibre of person; and as a group (so far) they’ve done a great job of letting the right people in (and leaving the rest to be jealous from afar).

We can do the same - and the stage4 tarballs give us a great opportunity to do so. We can use them to raise our standards where it matters - in how it affects our users. We move onwards, and the developers mentioned in Donnie’s post will either raise their game to match, or they’ll fall behind and be left only with their clique to keep them company (mmm, and that would be very ironic indeed). At the same time, raising the bar will attract new volunteers to Gentoo, folks who’ll surprise us and delight us with the great things that they’ll do. If we’re no longer wallowing in the brown stuff, we’ll stop attracting flies.

Whatever else you may think of this idea, it can’t be any less successful in dealing with our unwelcome members than just making them the most unpopular kids in the class, can it? Maybe they’ll go through with their threat of forking - and history shows just how dumb an idea that is around here.

Puppet is a Ruby-based alternative to the venerable cfengine. There’s a bug open for it in Bugzilla. Until I know whether it’s worth adding to the Portage tree or not, I’ve added it to my server overlay. You can add my overlay to your Gentoo install by running layman -a stuart-server.

I’m interested in testing Puppet as a tool for pushing configuration updates out to EC2 AMI’s. I’m also interested in modifying Puppet to be a two-way tool, so that when a Xen VM boots, it uses Puppet to register its IP address with a central server. This could prove a useful part in the puzzle of figuring out how to take advantage of EC2’s dynamic nature.

I did some (very basic) testing of Amazon’s EC2 tools on Gentoo last night. I had no trouble starting or stopping instances, generating keypairs, monitoring instances, or listing the images available on S3. Total cost of last night’s testing was 11 cents US.

Next step is to get a simple ebuild for the EC2 tools done and into Portage. So far, their only DEP is sun-jdk-1.5. I plan on putting them into /opt/ec2-api-tools, and the package will be called app-admin/ec2-api-tools.

Things Learned About EC2

• For computing on demand, EC2 is cheap.
• Although the docs say that it takes minutes to start a virtual machine, I thought they’d done the usual thing of just putting that in there to cover themselves; I thought (especially at the limited beta stage) that a machine would come up in under a minute. Boy was I wrong . I had time to go and make a fresh brew. EC2 looks fine for computing-on-demand in the traditional meaning … but it’ll be an interesting challenge to use it (say) for dynamically increasing webserver capacity.

Things Added To The TODO List

• Create the app-admin/ec2-api-tools package

I was going to expand on Donnie’s post tonight - honest - but I got sidetracked by the copy of openldap I found in the stage3-i686-2006.1 tarball I built my Xen image from (wtf is that doing in there?), and then by some great help from rocket on putting together a stage3 X86 for Xen. No promises, but we might just get a ready-built stage3 for Xen at some point.

Maybe in the morning.