Everything and the Mobile Software Universe…

Recently Samsung announced Bada, a new development environment: in a nutshell this is an SDK with a set of C/C++ API associated to an application framework, ported on top of Samsung legacy RTOS, or Linux.
Bada will be deployed in a majority of Samsung touch phones from smartphones to feature phones.
Antony has been quicker than me and posted a nice Bada article at Vision Mobile, depicting why this Samsung move may prove to be a wise one. I mostly agree with him, his arguments are around gross margins, market pressure and differentiation. I’ll try to dig a little more around this differentiation aspects in the Samsung case and why they really seem to innovate in this area, especially versus an Android strategy.

When Bada was announced, I was very negative, my first reactions and comments were harsh: again a me too initiative from Samsung, the OEM with fantastic execution but no clear vision of its services/software strategy, trying all the available OS on the planet and waiting to see if one is successful.
Then I’ve looked back at what they achieved recently and what they’ve announced with Bada.

• They are now the undisputed number 2 in number of devices sold (not far from 20% of the market in Q2 2009 for ex)
• The Touchwiz UI 3.0 is on the smartphones AND ALL the other touch phones of the company: difficult to tell who is or is not a smartphone now (I’ve made the experiment at the last CTIA).
• Release of the Touchwiz UI widget SDK to develop widgets for all the Touchwiz based phones.

What are they trying to do with this TouchWiz UI ?: use the best software platforms (RTOS for cost effectiveness and integration, high level OS for SDKs and features) while trying to uniformize the user experience and consolidate the Samsung brand.

Obviously the next step, to retain its customers, a politically way to say “lock them in”, is to have exclusive applications and services accross the Samsung devices line…even better if the customer has paid for it so he won’t throw away its application investments to buy a competitor phone for its next purchase. Same strategy as Apple, but here we are talking about Samsung a company with dozens if not hundreds of different device models, across all the price ranges, selling more than 200 millions phones a year (yes a year! how many iPhones sold today? , ok margins, blablabla…and yes I have an iPhone).
How to do that? Make a robust application environment, OS agnostic, with a dedicated SDK, to allow deployment of the same binary to a range of devices with different software platforms … well this is exactly the description of Bada.
We (speaking in the name of OpenPlug in this sentence only) have advocated this very same idea for the last 7 years to push OEMs in this direction (Samsung and Nokia were part of the lot ), and this is at the end taking of : Nokia with Qt (still has to deliver but the intention is clear) and Samsung with Bada.

But why creating a new one and not simply reuse Android code base?…I’m sure you have the answer already:

• Android is free? Android is by no extend free: you have to pay a lot of your R&D budget to make a phone with your brand, your services, etc…
• Android is open, why not getting it? NO Android is a closed box, Google and only Google can really change it, it’s really time for the industry to wake up: it’s not because you have the code of something that you can control it. If you don’t have the roadmap and the team who is maintaining and developing it you simply have meaningless mega bytes of symbols
• …but this is from Google, the good guys! Sorry, they are not, they are pushing their own services, not yours, and now with the NexusOne their own phone and user experience.

Google/Android has a complete opposite aim versus OEMs own agenda:

• Android is here to push Google services in the mobile world, as a corollary it allows you to swap your device as easily as possible because the new one will have all your data and applications loaded and compatible from the old one.
• … wait,wait this is exactly what an OEM doesn’t want: an OEM wants you to buy your next phone from them, not the competitor. So you have to differentiate.

Samsung, like Nokia, simply doesn’t want that an external third party software house decides if its devices has to be like that or like this… to be look alike brothers to its direct competitors (check the Windows Mobile phones for the last 5 years, and you will get the point : all the same).
So why not following their own path? Apple has made the choice and is successful, Nokia is trying, Samsung has to move, and for me it is doing so in a more pragmatic way than Nokia:

• Bada is not about reinventing a new OS with the underlying plumbing to the hardware as is Nokia Mameo, who cares of that today?
• The User Experience is already in production and refined device after device
• Bada SDK is not “fancy” but raw C/C++ … but what the point as long as I can sell and do my applications for hundred of millions of devices and customers (let the fun to the WebOS guys…I’ll get the cash)

Of course Bada is not quite their today, where are the devices? where is the market place? But it is refreshing to see Samsung taking its own path, its own direction…opening even more opportunities to application developers.

[UPDATED : Added a BootCamp test to compare it to VM solutions, changed the Parallels boot time by the numbers obtained with a fresh VM with al my setup in it.]

It’s done, I’ve switched to MacOS … but I still work and still lead the Elips Studio dev team, and (for now) the product is still Windows Only.
Yes, from time to time I need a window box to get some C/C++ work done. Bootcamp (ie booting winXP on a Mac) is not the best option for me as I’ve made the switch to have my mail, docs and all under MacOS.
So I’ve looked at a virtualization solution. I begun with VMWare Fusion 2 under Lepoard 1O.5.8 it really was OK , I’ve never looked at Parallels.

Here are some benchs I’ve made to configure my first VM:
For the exact same Build environment on my new top of the line MacBook Pro, 4GB RAM, 7200 rpm 500GB HDD, 3.06GHz Core 2 Duo CPU:

• My initial dell laptop (D630 2GB memory, 2GHz CPU):  5:40
• vm 2GB split no prealoc 8:20
• vm prealoc 40g :7:40
• vm prealoc 2GB split : 7:40

=> My old PC was 25% faster than my VM … well not that good but still usable on a daily basis.

Then I’ve switched to Snow Leopard, and I’m not sure but I really felt that my fusion VM began to slow down, I wasn’t able to conduct the same test to measure this degradation, so I’ve waited for VMWare Fusion 3 and Parallels 5 to fix it.

And here we are, they are both available:

The Bench setup:

• I’ve converted my initial Fusion 2 VM to Fusion V3 and Parallels v5.
• I’m building a Makefile Based Visual Studio Project of hundreds if not
  thousands of C and C++ files (the whole ELIPS Studio runtime)
• VMs configuration : 1.8 GB RAM, max performance settings, in fullsceen mode (no unity or coherence gadgets enabled), no shared stuffs
• Reboot of the MacBook between each tests
• UPDATED: added the same bench under Bootcamp

I was shocked by the awfully long boot time of my converted VM under parallels v5, so I’ve setup another VM with only XP in it, and … it boots pretty fast, so perhaps an issue with the conversion of my VM from VMWare to Parallels. The long phase is after the login, when XP is loading my “user preferences”, I’ll look at it more closely pretty soon. UPDATED: I now use a fresh VM, as seen in the number, no issues
Beside this:

• Fusion v3 seemed to be a nice improvement with its 64bits engine but  with only 7% improvement for the build compared to v2, well this is disappointing
• Parallels v5 is just in another league, 39% faster than fusion v2 and 34% (a third!) than v3 for build…I’m faster than on my former laptop!
• … and Parallels just “feel” faster: UI is slightly more responsive as launching apps.
• UPDATED: Bootcamp: build time under bootcamp are 50% better than Fusion (so I Build twice as fast ), for Parallels we still have a 22% improvement, around 6mn, with 6 or 7 of those full build a day it is a 40mn improvement of my productivity…ok not so big, as I’m not able to do a lot of stuff under Windows (all my documents, settings, etc are under MacOS).

Conclusion:

Well beside the boot time issue in Parallels (that I’m fixing using a “from scratch” VM), due to the speed increase in Visual I really can’t go back to VMWare for now.
I’ve just bought Parallels 5, and oh, they are offering a 30$ rebate if you have a VMware Fusion License (!! yes really, this is competition…), check it here, bottom right of the page.

UPDATED: After few days of heavy use, no stability issues with Parallels, The Mac is not slowed down when not compiling (as with Fusion), so really, for me Parallels is a superior product. I’m also experiencing Bootcamp : for sure it is really faster, but not so much compared to Parallels and I loose a lot in usability. The next test I’ll conduct is … building under a Bootcamp loaded as a VM in Parallels, perhaps the best of both world?

Other reviews and articles:

• Einar Ingebrigtsen has some nice tips to tweak your parallels VM … but my benchs are simply showing that plain disc and SCSI are not really faster.
• lostWhisper has a good head to head fusion v3 vs Parallels 5 comparison also: He prefers VMWare for dev… I don’t due to the size of our projects, I’m just more productive.
• The Mac  Village Blog has a great screencast,
  really helpfull to understand the two products, at the end of the
  screencast you have a startup bench, where parallels really faster than
  Fusion for startup (again I think my numbers for startup are biased)
• ATLChris is also giving its view on the two products…and recommends Parallels 5 also

Parallels 5 really seems to be a winner for them also