1) The momentum of asterisk
. In 2005 - cheap, good voip phones will become common
. SIP and IAX related IVR and call processing services like voxeo
will become more widely used. And entirely new levels
of voice and data integration will also appear. There are still barriers to entry, notably monopolist-wanna-be Scansoft has gone and bought up all the speech to text software in the known universe
, and the VoxML standard needs a good Linux library, but I have confidence that these problems will begin to be resolved, as well, in 2005. Watch for big plays in Asterisk by major vendors in the latter half of the year.
2) This will be the year that Linux 2.6 finally becomes stable, and the enhanced features of the OS, especially on 64 bit and embedded platforms, will entrench Linux in two markets that MicroSuck had hoped to dominate with the oft-delayed release of Longhorn.
3) Longhorn won't ship this year.
4) Somebody will ship a 8 way Opteron this year. Few will buy it.
Networking - Viva IPv6!
1) With China leading the way
, IPv6 will finally begin to gain a foothold with established ISPs. I never would have guessed that in 1996, when I first started playing with IPv6, that it would not have rolled out by now. I thought everybody would have a dedicated IP address, not this dynamic DHCP driven cyber-serfness we have now. I don't expect IPv6 to be rolled out in the US for at least another 3 years, but I think the ball is finally starting to roll, driven by the needs of one populous country.
IPv6 needs to happen to finish connecting up the world. Let's make it happen! My christmas wish is for everyone with a dynamic IP address to bug their provider for a real, static ipv6 one.
Hardware - the Up and coming System On Chip - the Cirrus Maverick EP9302
Linux, running on the Cirrus Logic EP9302
, will find its way into new and amazing markets. It will displace hybrid arm/DSP designs such as the TI-OMAP, and wedge into places where only the mips has traveled before.
1) The EP9302 has a real FPU, (not some bletcherous grafted on DSP part that requires you program it in kernel space). It "just works". For echo cancelation it outperforms a 533 mhz Xscale by a factor of 10. MP3 playback and encoding have similar ratios. Best of all - the part is less than 10 bucks.
2) It's a true host processor, supporting host USB mode. This means that standard USB peripherals "Just work". (IMHO Intel has been intentionally crippling the Xscale line keeping it as a slave processor, and leaving off an FPU, because it would compete with their x86 bread and butter.)
3) Have I bitched about DSPs enough? DSPs are great - IF you don't have a MMU to deal with and your code complexity is limited. Otherwise, adding a DSP to your embedded design always throws a monkey wrench in your schedule. The EP9302 floating point unit is transparent to normal software and fast enough vs most DSP needs.
If you are a embedded system developer with DSP problems, get an EP9302 development board
, get the compiler
, and a kernel, and throw your code at it.
I've spent a long couple years looking for a replacement to the StrongArm. I think this is it. I wish it clocked in at 500Mhz, but you can't have everything.
I hope more minds expand... towards helping the private space programs truly take off