* Scratch resistant screen! Preferably resistive, but if that is not possible, then let it be capacitive.
Dual mode touch screen like in tablet PC ? Able to use finger touch as well as more precise magnetic stylus. We'll even have cursor hover feature instead of just clicking.
I totally disagree. There should be NO screen outside the visible screen - NO BEZEL! It's a waste of space and makes the handset too big to carry. The N900 could have a 5" screen with the current form factor.
...
* Enlarge the touch screen outside the visible screen. Perhaps even on the sides of the unit.
...
That's a GREAT idea. I wonder if it could substitute for a D-Pad.
I totally disagree. There should be NO screen outside the visible screen - NO BEZEL! It's a waste of space and makes the handset too big to carry. The N900 could have a 5" screen with the current form factor.
That would be cool, but where would the camera and proximity sensor be?
When did phones get hardware floating point math?
(GPU and DSP don't count unless math functions in the standard libraries support them out of the box.)
Actually, I wouldn't mind if floating point processing went the way of the dinosaur. 64-bit integers are faster, have greater range (in the sense of more states) and don't have the many rounding-related problems that floating-point math has.
It does. The core of Maemo will cease sending events and wakeups to applications. Misbehaved applications can still screw things up, and usually hildon-desktop screwing up is the result of a poorly written plugin.
It can't -force- applications to stop, but it easily could be configured to do so.
But generic PC has a disk for swap with 50-100MByte/sec write speed and N900 - flash with 19MB/sec write speed.
No PC ever hits more than a couple hundred KBytes/sec when being forced to page. Depending on the flash it can be way faster, but swap is still ungodly slow compared to RAM.
So yes, more RAM. Always more RAM. Paging is the symptom of a problem, not a solution.
It does. The core of Maemo will cease sending events and wakeups to applications. Misbehaved applications can still screw things up, and usually hildon-desktop screwing up is the result of a poorly written plugin.
I think it has sense to clarify here - what do you mean with "core of Maemo". I did my own event logging and I know for sure that there are a lot of applications which don't cease an activity with locked screen. In my surprise I even see a frequent polling of orientation by MCE with screen locked and even without ANY application which requests this info. Damn!
Originally Posted by
It can't -force- applications to stop, but it easily could be configured to do so.
I can do it. Kernel has mechanics - "cgroups". It is only need to be configured to have a right cgroup and I compiled my own kernel with it and tested. But I can't stop MCE.
Originally Posted by
No PC ever hits more than a couple hundred KBytes/sec when being forced to page.
You mixed paging with swapping. Both uses the same mechanics but swapping means - "swap out the whole application" and that enjoys the sequential disk writes with 50MB/sec bandwidth in contrast with random paging. You can measure it yourself, BTW.
"Easily swappable Battery ("Hot Swappable" a bonus)"
Amen to this suggestion, but hot swappable is a must, not a bonus. It was suggested for the N900 but ignored. I predict Nokia will ignore the suggestion this time also. It's too bad, because in these days before batteries get really capacious, hot swapping would be a fantabulous solution.
(BTW: I find the N900 battery quite easily swappable already.)
Swapping hasn't meant "write out the entire application to storage" since before www was invented...
Today when the kernel needs to free up ram, it has among other things Least Recently Used lists, whether the page is disk cache, dirty pages waiting to be written to storage, executable pages belonging to apps and libraries that can be just dropped from memory (and read back from the fie in question later if it's needed), or anonymous pages (pages that don't already exist on storage) that get pushed out to swap, doesn't make much difference...
Yes in the old days swapping meant writing reading entire processes from RAM, but these days it's just on-demand pagein/pageout of anonymous memory to a swap area..
