Archive for June 14th, 2006

Mobile memory hogs

At last, the Minimo guys fixed their bug with the Axim x50v and so version 0.016 now loads correctly on my PDA. I did a few tests with it, and apparently Minimo requires 14 MBs of RAM to do anything useful on the web. More complex sites, like CNN.com require about 18+ MBs of available RAM to render. And if you use more than one tabs, you will need at least 24 MBs of RAM to operate correctly.

Now, the x50v and most PDAs come with 64MBs of RAM, but on Win2003SE system software (like the .NET framework, or drivers) must be installed on the RAM. This means that a well-equipped PDA (software-wise), has about 25 MBs of RAM free after a clean boot. WinMobile5 does not use the RAM to store system software anymore, but it uses large portions of it for caching. Depending on how clueful the OEM was (usually they are not) and especially if the device is also a phone, you are ending up with LESS free RAM than you would be with a Win2003SE PDA. My Qtek 9100 WM5 phone for example only has 24 MBs of free RAM after a clean boot.

This means that Minimo will barely work on today’s most powerful PDAs. Now, think that non-Windows smartphones (e.g. Symbian S60 and UIQ) are using many times less powerful hardware: usually at 220 Mhz with 32 or 16 MBs of RAM altogether. Same goes for Palm devices which while they have lots of RAM-storage, they don’t have more than 16 MBs of “dynamic RAM” allocated. This means that Minimo simply can not run on these devices. At least, not until about 2010 or so.

The reason I write all this is because it’s really pissing me off when I write somewhere about a mobile browser and what not, and then you have a Linux wheenie spring out like little fart proclaiming “…but, don’t forget Minimo!!”. Sure, how can we forget Minimo? It’s unforgettable. Our device felt like it had anal sex with a 15cm stick when trying to fit it all in its RAM.

Minimo

Now, to be fair, Opera and the new S60 Safari/KHTML browser require quite some memory too: about 5 MBs to load to RAM initially, and about 9 MBs each to render a complex site like CNN. If this doesn’t sound like too much to you, remember that companies like Teleca and Openwave have browsers that are loading in 400 KBs of RAM and render CNN in 4 MBs resulting in cheaper devices (less CPU and memory required).

No, Minimo is not bloated per se. It’s just that it was never architected for mobile devices. It was designed for desktops with lots of RAM. Trying to make a big browser like Mozilla fit on a mobile device’s constraints requires lots of re-architecture, which is not what the Minimo guys are interested in because it’s a lot of work, and because it requires huge rewrites in some places. When they do that, they would be probably losing compatibility with the main Firefox source tree too, making their work even more difficult. So, instead of re-architecting, they are waiting by fixing only a few things. They are waiting a time when devices will come by default with 128 MBs of RAM. In about 4-5 years that is.

Regardless, memory is not the only problem of Minimo. Speed is too. I have one of the fastest PDAs ever, a 624 Mhz monster, and even that can’t make Minimo to feel responsive (takes seconds for redraws of simple menus). Which is pretty weird, as my 333 Mhz PII-mobile 7-year laptop runs Firefox way faster than Minimo generally runs on my fast PDA (although I am familiar with ARM not being as fast as x86 at the same Mhz, and PDAs not using fast RAM and enough cache).

Why Apple snubs its open source geeks

Tom Yager argues how unfair is for Apple to close their kernel’s source after being open for years.

You all know that I am no one’s fangirl, but I will have to stand beside Apple on this one: Apple made a concious choice in 1998 to use FreeBSD’s code to enhance NeXTSTEP’s own Mach kernel and create Mac OS X. If Apple didn’t want to have a business choice of closing their code at some time in the future, they would have gone with Linux or QNX. But no, they had consiously chosen FreeBSD and the BSD license, which is a business-friendly license.

“But, but, but what about the coders?” I hear… Well, what about them? Apple is a business first and foremost and then a charity or a bolster of democracy. They have an obligation to their shareholders before anyone else. And if the closing of their kernel means less piracy in the x86 world, then so be it.

It sucks for some open source developers, but let’s face it: very few outside of Apple enhanced that code or had a practical use for it. Instead, that Intel-based open code was used more to HACK it and PIRATE OSX than to actually ENHANCE it. It is on the best interest of Apple and its shareholders for this code to be closed until Apple actually uses a DRM chip on their hardware that can’t be easily hacked or emulated.

I just hope that some academic and high-end projects can get that code if required, after a special license. Sun Solaris was licensed under such terms and its full source code was given away to companies (long before it became officially open source) that needed to modify it for their purposes, for the right price: $100,000.