Firefox 4: One generation ahead of everyone else speedwise

Mozilla's Chris Blizzard in an interview about the rising competition by Google Chrome, the evolution of the web platform and the prospects for WebM

Over the last few years Firefox enjoyed the role of the rising star in the browser world. Nowadays Google Chrome seems to have taken that very spot, currently even grabbing some market share from Mozilla. During the Linux desktop conference GUADEC Andreas Proschofsky had the chance to sit down with Chris Blizzard, "Director of Web Platform" at Mozilla and conduct the following interview about the current situation in the market, the advancements of the web as a platform and upcoming improvements for Firefox 4+.

derStandard.at: When we talked last year you observed that although Google Chrome is gaining a lot of publicity, most of the developers at tech-conferences - like GUADEC - still use Firefox. If I take a look around this year, this seems to have changed considerably, a lot of those early adopters seem to have moved to Chrome / Chromium. Do you have a problem in that specific - but influential - user segment?

Chris Blizzard: Actually I think a lot of people are using both now, but it's interesting to see for sure. From a web developer standpoint I still think Firefox' tools are far superior than anybody else's.

Interestingly we haven't seen our user numbers change that much, even though Chrome is having big wins in this segment. The thing is all those numbers reported are actually usage numbers and not user numbers, so early adopters - who use the web heavily - influence those a lot more than others. So it looks like Chrome is used by more people than it actually is. The interesting thing is, we had the same effect in our early Firefox times, we just didn't realize it back then as we didn't have the proper tools for that.

We've also done some research and what is interesting is that a lot of people seem to choose their browser on very specific features, so for instance a lot of people who prefer to see their most recently used sites when they open their browser, they use Chrome. If they want a clean blank page, they use Firefox instead. So I think what we're starting to see is, different people want different experiences, and are going to choose their browser based on that preferences. And we're just seeing the first shift in that direction.

We know we're not going to be able to serve everyones needs and that's not even our goal as an organization. We want to drive the web as a platform and we've been pretty successful at that.

derStandard.at: Google is now changing to a cycle where they offer a new stable feature update for Chrome every six weeks, trying to push innovations faster to the users. Mozilla on the other hand roughly releases once a year, doesn't that put you into a disadvantage marketing-wise?

Chris Blizzard: Well it depends on which pace you want to have. The Chrome guys are approaching that a bit differently than we are. It'll be interesting to see if anybody else than the early adopters are going to be okay with their browser changing every month and a half. We prefer to take more time to prepare people to bigger interface changes.

Still we don't always change the version number for big changes, for instance the out-of-process-plugins we recently introduced to the 3.6.x-series was a giant amount of work and it really improved peoples experience with Firefox.

I'm actually a little bit skeptical about a six week cycle, where do you find the time to really innovate in such a short time span? But going faster is something that we definitely would like to do too, we just have to figure out the right pace for us.

derStandard.at: One of the advantages of fast release cycle is that Google is able to push things like the upcoming Chrome web store to their users pretty fast. Is such a store something you'd also like to do at Mozilla?

Chris Blizzard: Personally I think an app store for the web is a pretty interesting idea, basically making web apps more easily accessible and also providing a monetization model. But it has to be done in context with a lot of other things to get it right, so we're going to concentrate on that first.

derStandard.at: What parts are still missing for a good web store?

Chris Blizzard: So Javascript performance is really important, but we are getting there, we're going to talk about near-native performance in the future. Then there's WebGL for 3D-applications, we also have to fix offline storage which is quite a mess at the moment. Also you have to be very cautious about security and we are working on some of that stuff right now. So for instance with ForceTLS a website will be able to tell a browser to only talk to it encrypted, which prevents certain types of attacks. We're also working on content security policy, which is going to help against a whole class of cross-site-scripting-attacks.

derStandard.at: In recent Javascript benchmarks Firefox seems to have dropped to the lower end of the spectrum, even being surpassed by Internet Explorer 9, are you loosing track here?

Chris Blizzard: Firefox 4 will bring a lot of improvements in this area too, they're just not in the current development builds yet. What we've seen is that in places where our Tracing-Engine gets used we are actually faster than anyone else, it's just in those cases where it doesn't fit that others do a lot better. So we're trying to improve our baseline performance and combine that with the Tracing-JIT, with this we'll be one generation ahead of everyone else.

But the real story here is, that there are still a huge amount of opportunities for improvements. The current generation of Javascript-Engines has plateaued, I mean the percentage difference we are seeing between all the current browsers at the moment is not very big anyway. Also the benchmarks are not very useful anymore, as those number are mostly not influenced by real Javascript performance anymore. Just to give you an example of this negative impact of benchmarks: We have to do stuff like optimizing Daylight saving time lookups because that is influencing some benchmarks negatively, so we're not doing actual Javascript improvements here. Sunspider has these problems, V8 also consists of some crazy code, so it's hard to find some good benchmarks.

derStandard.at: Firefox 4 is going to use hardware acceleration through Direct2D and DirectWrite on Windows, are similar things coming up for Linux and Mac OS X?

Chris Blizzard: Within what's provided: Yes. We're trying to give the best experience possible on each platform. So for Windows Vista and 7 we see huge improvements when doing certain graphically intensive stuff. On OS X for example we have support for OpenGL for doing compositing, on Linux we do the same. But generally the Windows APIs that we have are better and more rich than what we have on other platforms. To give you an example: On Linux Cairo and Pixman were supposed to be fast, but unfortunately the underlying infrastructure never really got fast. On OS X we are actually pretty fast but Direct2D gives the performance advantage to Windows at the moment.

derStandard.at: The out-of-process plugins were supposed to be only the first step of a multi-process Firefox, having different tabs in separate processes. That's one notable admission in the plans for Firefox 4, is this still coming up?

Chris Blizzard: Yes, sure - though not in Firefox 4. We might actually see it in the mobile release before it gets into the desktop, which is pretty interesting, it just happens to be easier there.

derStandard.at: Recently you took over the role of being the "Director of Web Platform" for Mozilla, how does that change your daily responsibilities?

Chris Blizzard: Well my old job was mostly about running the evangelization and the public relations groups, but I've moved on to do more and more platform work, and all that stuff - thinking about the future of the web, talking to other browser vendors on a regular basis - turned out to be really important, so I decided to try and focus on that full time.

For instance I just attended an IETF meeting, where we had some good discussions about http, about web-sockets and also about some codec-stuff. They're building a next generation audio codec for real-time-communications and specifying through the IETF. It's basically a combination of existing technologies with a focus on very low latency, where things like Vorbis or MP3 are typically bad. This is some work that gets done in cooperation with Skype.

derStandard.at: One of the big announcements of the last few months was the creation of WebM as a new open video format based on VP8 and Vorbis. Do you think it has better chances to be successful than Theora?

Chris Blizzard: Yes - because of the quality issues. To give some background: The interesting thing is, when you are doing H.264 you basically have to choose one of the three profiles that are part of H.264. And usually video site have to do "baseline", because the iPhone only does "baseline", at least older versions of which there are tens of millions users out there. So when you compare video formats you really have to compare H.264 "baseline" to VP8 and this is not just a rhetorical trick, it's actually the reality for all services. So for instance I talked to Youtube and the told me they tried to use another H.264 profile but had to go back to "baseline" because of the lack of support. And on that base the quality issue is no issue anymore, VP8 is absolutely competitive here.

If you talk about browser support for HTML5 video you also really need to look at timelines and trends in the market. If you think about a world where we have HTML5 video universally available in browsers - by that I mean more than 80 percent which should be something like two years away - you have to look at browser trends and shifts in operating system market share. And what you then realize is that you basically end up with a split market. So if you want to deliver video you'll have to support both H.264 and WebM. In the longer term I'm pretty confident that the free codecs are better aligned for the future, especially if you take upcoming trends like real-time-communications or mobile platforms in consideration, where WebM is really good - or is going to be because of partnerships with hardware vendors.

derStandard.at: Firefox 4 introduces the ability to build "light-weight" add-ons in form of Jetpacks, when are those going to replace the current generation of add-ons?

Chris Blizzard: There not meant to replace XUL-based add-ons. Firefox is a browser platform which is extremely extensible across a broad range of interfaces, you can touch a lot of things inside the browser. That's something no other browser allows. And as a project this is amazing, a huge percentage of innovation has happened in extensions before it happened in the browser itself.

Jetpacks for us are basically an equivalent to what Chrome Extensions provides, which is a lot less than what you can do with our add-ons. But they will be a lot easier to build,
we are going to provide a web-based editor which should lower the barrier for extension development. So Jetpacks are really easy to build, really easy to test and you can install them without restarting the browser. By the way you can also do that with XUL-based add-ons in Firefox 4, it depends on the add-on and it has to be carefully written and actually announce it, but you can now also install them without restarting.

(Andreas Proschofsky, derStandard.at, 18.08.10)

Share if you care