I have a SageTV HD Theater, dual tuner HD HomeRun, a D-Link NAS and an original Mac Mini running SageTV Server. This provides two major benefits:

* ability to stream audio, video and photos from a computer or NAS to the TV
* ability to record and playback HD over-the-air broadcasts with DVR software

The living room set up is:

HD OTA antenna ---> HD HomeRun ---> GB Switch AV Receiver and TV

Then I've got a cat 6 ethernet cable running under the house connecting the switch to a router in another room:

D-Link NAS --> router --- Mac Mini

With respect to noise, this works well, since the HD Homerun and SageTV are completely silent (though the blinking green network activity LED on the SageTV should also be silenced).

I kept costs down by using a free, first generation Mac Mini that was headed for the electronics graveyard. But, it appears to be too under performing for the task. I had to upgrade it to 1 GB of RAM for it to be tolerable, but I'm still having problems when using it with the NAS.

The setup does actually work well as long as it is writing just to the internal hard disk. If I have it write to the NAS and then stream from the NAS through the Mini, most HD channels have an intolerable amount of stutter. However, I still need to experiment with directly mounting the NAS from the SageTV and using an external USB drive on the Mini. I also purchased the Placeshifter software, but haven't had time to try it out, yet.

I wrote a lot of Occam code for transputers around 1989-1992. I thought the language was well-designed, certainly for that time period, and very simple to learn. I was able to pick up all the syntax and start coding productively in less than an afternoon, although I already did know C. The message passing syntax is especially simple. The complete ABNF grammar is only a couple of pages.

We built a parallel pipeline for signal processing. Adding more processors was as simple as connecting a few pins, modifying our config file and restarting the system. I usually ran it with 48 transputers in the grid. The same Occam code ran on each chip, though there was some special case code executed only by the chip taking in data from an array of Motorola DSP chips and the chip combining all the processed data to hand off to a 486 for display. The algorithm was effectively a pipelined map reduce with real-time constraints.

Debugging was a bit challenging. We resorted to two approaches - hooking up LEDs to unused pins and raising a signal on a pin (obviously good for only simple stuff) and writing code to pass log messages down the pipeline that got funneled to one chip for display on a console.

Waves Audio is offering a processing chain for their own tool as well as for Cubase and the very popular ProTools.

http://www.waves.com/content.aspx?id=5798

The equalizer is $300 and the noise supressor is $2900. I think they are targeting above my level of concern.

A few years ago someone opened 9 credit card accounts using my SSN and managed to run up US$13,000 in charges before all the accounts were frozen. In addition to changing my address and birthdate with the three major credit reporting agencies, he altered the name he used to open new accounts, almost as if he wanted to see how far he could push it.

After using my real name for a couple of accounts, he opened an account using Bobby as a first name, instead of Robert. Sure, it's a common nickname, but it appeared no where else before that point on my credit reports. And then he switched to Bobbie. Next he opened an account with a different middle initial of L. As the coup de grace (or just a careless mistake), he reversed the first and last names and opened an account as Stewart L. Bobbie. I don't know if any of his applications were rejected, but over half of the 9 had significantly incorrect information and were still approved. In addition to the name changes, he used three different addresses in the midwest in states where I had never lived.

For the first couple of accounts, he made a few initial payments before stopping payment. I mentioned this to my brother-in-law who is a police officer who had previously specialized in solving economic crimes. He said this was a common strategy of the savvier fraudsters. After making a few mid-size charges and payments (often returning the merchandise where they can get cash rather than credit back), they go for a much bigger charge. If they went for the bigger charge right off the bat, it's more likely it would have been rejected and the account frozen.

--single-transaction and --skip-lock-tables are mutually exclusive, so you don't need to include --skip-lock-tables. Also, --opt is added by default, so you don't need it, either. Of course, it doesn't hurt to include them, other than making the command longer.

Best info I could find quickly was an interview at robotsrule.com with John Sosoka, CTO of Ugobe. Here's a couple of the more relevant quotes:

"Pleo does have his USB port and an SD card to allow Pleo to be updated."

"In order to play with voice recognition you have to write it in C or assembly code. While possible, it won't be supported by the PDK tools at first, although we won't do anything to stop you. We want to get the controls in to let people to quickly be able to play with the sensors and motors, and things like that. Over time we will create better tools for people to do lower level things with Pleo."

"People will have brilliant ideas of what to do. We will allow them to implement them. We want to see what can be done."

"We actually started with some proprietary software at first and one of the reasons I wanted to make the switch was to allow programmers to just get gcc and some other tools and you're good to go. You can run gcc and just start working your way through it."

"In terms of Pleo's internal memory, the main memory in Pleo is 4 megabytes. I think Pleo is in the 50 Megahertz range for the main processor and 64 Megahertz for the processor in the back."