Česká verze rozhovoru: Rozhovor s Denise Dumas z Red Hatu o budoucnosti Fedory a spol.
You’ve been heard to say that there will be classic GNOME 3 mode in Red Hat Enterprise Linux7 – is that one of the reasons you chose to attend GUADEC? To steer things in the right direction?
No, I think control is an illusion. This is open source [laughs]. Actually, I’m here because a lot of the Red Hat desktop team gathers at GUADEC each year, so it’s a great chance for me to see everybody at once because our teams are so globally dispersed that it’s hard to get everybody in the same place at the same time. We have a big desktop team in Westford, and we have a big desktop team in Brno, and we have lots of people who work out of their homes all over the world. This is a chance to see everybody at once, and also a chance to see what’s going on in the community.
I think GUADEC is a great example of a community that works well together, and it’s always good to see the way that they function.
You recently, at the Red Hat Summit, introduced the Red Hat Enterprise Linux Roadmap. With regard to that, how is Fedora doing? Fedora is the oft-cited staging ground for Red Hat Enterprise Linux, so does it live up to the expectations, does it satisfy your requirements, or is there anything that needs some serious tweaking?
Flock is next week, so I’m going to be there, and a number of Red Hat people are going as well. You’ve probably seen the list of proposals, and it gets very interesting. It’s not like the Red Hat people would say “here’s what we’re going to do, and our goals for Fedora are...”. I think we do have a few things that we think Fedora can do better. I think Matt Miller’s proposal about the rings is very interesting.
There’s been so much controversy about how do you keep Fedora competitive, and how do you attract users and developers to Fedora. I’m hoping that if we can get some agreement on what the direction is going to look like and get people moving in a consistent direction, we can get some agreement about what we need to do. If I can figure out where they want to go, then maybe I can help with the direction along, I can help Red Hat people get some free time to contribute.
You mentioned Matt Miller’s proposal, Fedora.Next, which is what he will be discussing at Flock about a week from now, what is your take on that? It is, after all, quite dramatic.
Drama’s good. A little controversy...
Does Fedora – and down the road, Red Hat Enterprise Linux – need a major reorganization in terms of the way it is built and put together?
I think it makes sense. In an operating system, there always is the concept of what is core, what is the smallest footprint that you can deliver, and still call yourself complete. And Matt’s proposal, I think, tries to define what that is for Fedora, and open the possibility for additional layers of things that are able to move at a different pace, or are able to move in less constricted ways that give people more freedom.
Yeah, it’s “core”, but it’s not the same old Fedora Core and Extras model, I think. It’s “a core”, and then additional layers where we have the freedom to do different kinds of innovations. One of the big issues that Fedora’s always had is that it moves so fast. And it’s hard for people to keep up with. So, there’s actually another interesting proposal for a more stable rawhide, I think that’s Steve Gallagher’s. We have to keep it moving forward, but we also have to make it possible for people to use it more effectively. Because that’s the way you get people into the community.
Software collections are a pretty exciting development, and I believe it’s in the same vein – it enables people to use the new stuff without sacrificing the stable core or base. What is – right now – the most exciting new development in the Red Hat Enterprise Linux/Fedora world? Software collections come to mind, but what would you pick?
I agree with the software collections. Red Hat Enterprise Linux customers have been very positive. This is a problem that they’ve had. Software collections were controversial for Fedora, but I would love to see them accept software collections because in a weird way it fixes a Fedora problem too. With Fedora, you always have the most recent version of everything, but if you want to stabilize on an older version, that’s really hard for you to do. And that’s something that software collections can be good for for Fedora.
I think some of the container work is interesting. Of course, systemd. Systemd is always interesting. Lennart [Poettering] is breaking new ground, as always. Bless his heart. I think he’s gotten a lot better at doing it in ways that people can accept more easily. He’s gotten a lot more skilled in figuring out how to ease people’s fears about new technologies. There’s always that: I’ve got to have the newest, and oh my god, I can’t break Linux traditions, I can’t break stability, I can’t break applications. It’s a creative tension – we’ve got to find a way forward, and still manage to hold the old applications, the odd enterprise-class things. Hence, classic mode.
You can’t demand that people change. People change, or not. They change when they see that it benefits them. So you have to give them a way to do something that’s comfortable for them, give them the opportunity, lower the bar, and give them a chance to change. I want people to try modern mode, but I can’t insist that they do. But if it’s there, and it’s attractive, and they can still get their work done in the meantime because they don’t have to retrain their fingers, then there’s a much better chance that they’re going to adopt it.
So, in the long-term, you’d actually like the Red Hat Enterprise Linux desktop to adopt the modern GNOME?
I can’t really see that far down the road. I think that with Red Hat Enterprise Linux 7 our customers are going to have the option of the new mode for the first time, and we’ll see what they want. People vote with their feet. Our customer base tends to be conservative, but maybe it is changing as well. Everybody changes with the times. You just can’t dictate that people change.
You’re also bringing MariaDB to Red Hat Enterprise Linux, which many people think is pretty exciting. It almost seems a bit tongue-in-cheek...
...what? To say that we’re conservative and then consider MySQL alternatives, including MariaDB?
Considering that you’re looking at making XFS the default filesystems for boot, root, and user data. Are there any other reasons for this decision besides technical or engineering concerns? MySQL is an Oracle thing, and Chris Mason, the lead developer of Btrfs, used to work for Oracle. Is there any connection?
That’s interesting. Interesting that that’s what you would read into these decisions because really there isn’t. Yes, Oracle is the upstream for MySQL, and it’s difficult for us to support something for ten years when the way that they deal with security vulnerabilities means they don’t make it easy to do patches. But actually, no matter what, I think that Maria is an attractive option. They preserve the compatibility, it’s a really easy migration path. And they’re just great people. Monty’s [Widenius] wonderful, they’re so easy to work with, and I think that where they’re taking Maria is going to be a really good place. If you look at the extra connectors, and the extra things that they’re doing, the performance improvements... those are all features that make MariaDB worth a strong look.
And XFS versus Btrfs. We think that XFS is more suitable for enterprise customers. Some of the decisions that Btrfs has made in terms of their design make it maybe not so suitable for the kind of enterprise-class customer. Btrfs will be there, and it’s supported. It’s in Red Hat Enterprise Linux 6, even though it’s tech-preview there. Clearly, it’s available, and it will go full-support for Red Hat Enterprise Linux 7, but customers will pick what they want in the long run. I truly think that a lot of our customers are sufficiently conservative that they’re going to do an update in place and stay with Ext4. That wouldn’t shock me at all. I’ll watch to see what they decide.
You mentioned update-in-place, and I understand that one of the things you’ve been working on is support for smoother updates for Red Hat Enterprise Linux – without the need to do a complete reinstall between major releases. Fedora has come a long way in this regard – upgrades have been pretty smooth since the introduction of the fedup tool...
I love the name of that tool!
... Are you happy with that development?
You know, I was surprised that so many of our customers were interested in doing update-in-place. Clearly, there are disadvantages to that. You don’t get to choose the filesystems, you can’t take advantage of some of the newer technologies. But since that’s what they want, and since fedup exists in Fedora, and it is adaptable, we are looking at what would be possible with RHEL.
It’s hard when you’re dealing with applications because, of course, from the operating system perspective, we have no control over update paths for applications, so there’ll be some scenarios where we’ll do the best we can, and we’ll warn you ahead of time if it’s looks like this is going to be a risky path. You may have some big, complicated application and its altered config files, and maybe more things that we don’t know about. What we can do in that case is say, well, look, there’ve been all these alterations, so this is going to be a risky thing to update in place. We’ll try it for you, and here are the places where we think you’re going to want to look after we do this update to make sure things succeeded.
Do you think that this could, down the road, turn into some sort of automated set-up or upgrade? Is that something that you’re looking into?
Oh, in my dreams. I mean, I think our customers would love that. I think that’s certainly an area that would be attractive to our customer base. How difficult it would be to do something like that...? But who knows? We’ve got a few years till Red Hat Enterprise Linux 8.
The Brno office has recently launched an outreach program to get more women involved in programming roles. And since I’ve seen you referred to as a “he” on the web, I was wondering whether you’re a part of these efforts.
Well, Marina [Zhurakhinskaya] has, of course, participated so much in that, in the GNOME program. Clearly, we’ve provided funding and support, so that she was able to do a lot of that. I think it’s fabulous. I look around this conference, and I see all the women’s faces and think, what a difference this is from some other conferences. And a lot of that is the result of her effort. I’m going to go to the women’s dinner tonight, and I want to meet more of the people who’ve been involved in the effort.
I think attracting more women to open source is a really, really important goal.
How is it working out within Red Hat? All we know is from some PR mentions...
I think that fact that I’m here in the position that I am says good things for Red Hat, right? I started as a developer years ago. There are a number of women in Red Hat who are working to try to improve things for women. It’s a hard problem, and I think it’s actually become harder for a woman to be a developer.
How so?
Well, when I started – and this tells you how old I am, right? – when I started writing code, we didn’t know that women weren’t supposed to write code.
It’s great to have a brilliant idea, but you also have to be able to build a community around it. And in some ways that’s the harder part. Hopefully that’s Red Hat’s strength. We try really hard to make sure that we do things out in the open and with the communities – with, you know, varying degrees of success – but it’s always one of our corporate goals to do it out in the open and to build a community around it. Because things don’t prosper if it’s just one company in the upstream.
I believe Red Hat is doing pretty well in that regard.
I hope. We need to continue that tradition and keep doing the outreach and bring more people to open source and more women.