1. It works. I was actually doubtful that this thing worked at all. I can attest that its a real product, and it actually works.
2. It does connect to lots of different online photo sites, including Gallery2 and Picasaweb. Thats nice.

And now for the bad & the ugly:

1. The first card I got was broken, and diagnosing this was quite difficult. The replacement card seems to work okay.
2. I don’t actually understand the use case for this thing — you need to put the card in your laptop to associate it with a new wireless network. There’s no “use any open wireless network you can find” option. Thats a missing ‘killer’ feature, in my opinion
3. It won’t connect to the Google Wireless in Mountain View, or any other wireless network which requires an HTML login page.
4. Its slow. You have to leave the camera on for several minutes after taking a picture to wait for them to be uploaded to the web.
5. It doesn’t support movies. If you take a movie on the camera, it just sits there, and isn’t automatically uploaded.
6. It chews up the camera’s batteries. Yes, they say its low power, but plain flash memory uses literally no memory when the camera is idle. The Eye-Fi makes a noticable dent in the SD550′s battery life
7. It gets hot. The area near the card is noticibly warm when using the camera now. Thats sorta spooky, considering that Wi-Fi is nearly microwave frequency
8. The “Eye-Fi Manager” software is a web application. (why?!) You have to log in with userid & password if you want to change settings on the card. This makes absolutely no sense to me. I bought a piece of hardware, why does it require a connection to the ‘net just to configure it? Stoopid! Oh, and by the way, the Eye-Fi manager really seems like alpha-quality software anyway.
9. There’s no visible status indicator for the card. From the camera, there’s no way to see if all your pictures have been uploaded to the web.
10. It only comes in SD formfactor, not in CompactFlash. I’ve seen adapters, but they seem to be made out of metal, so I can imagine that they’d work very well

How this thing should work:

Since there’s clearly a reasonable CPU on the card, it should be doing something a lot smarter. One option is to put a bunch of ‘virtual’ pictures on the card that can be looked at and manipulate the settings of the card. For example, there should be a visible image that says something like:

“34 pictures taken so far, 19 uploaded to the web, 15 remaining. Currently associated with network “FooBar” with a speed of 3Mbps. Pictures will be posted to Picasa”

All that would be rendered in a JPG file that I could view using the playback interface of the camera. When it comes to configuration, the only operation that I could think of that every camera supports is ‘delete’. So, there could be a bunch of pictures, each of which says something like this:

Delete this photo to disable Wi-Fi connection for the card

or

Delete this photo to associate with open wireless network “Hacker’s Paradise”

The embedded CPU could realize when the photos were deleted, and then perform the appropriate operation. These 2 things: Turing the WiFi capability on and off, and associating with any open network, are absolutely critical to making this a reasonable and useful product.

My end verdict: Don’t buy the Eye-Fi card. It has such a limited use-case that its nearly useless. Its great for putting pictures taken at home on the web, live, and thats about it. Maybe that meets a lot of people’s needs, but it seems like its just not that useful for a broad audience. Its much much faster to just plug the card or camera into a USB 2.0 port and copy the pictures that way. Posting to the web automatically is really nice, but could be written as a software-only solution for anything connected via USB.

I came up with the following script that does a simple 2-pass encoding to produce a reasonable output file at 1.25Mbps. I choose to go with a higher bitrate and higher quality, but this could be turned down to around 700kbps and still be reasonable. Here’s the script:

#!/bin/bash
input=$1
output=./flv/${input/.avi/.flv}
logdir=`dirname $output`
mkdir -p $logdir
logfile=$output.log
ffmpeg -y -i $input -ar 11025 -ab 64k -aic -umv -b 1250k -r 30 -pass 1 $output -passlogfile $logfile
ffmpeg -y -i $input -ar 11025 -ab 64k -aic -umv -b 1250k -r 30 -pass 2 $output -passlogfile $logfile
rm $logfile*

And I run it like this:

# find . -iname “*.avi” | xargs -n1 -P2 ~/bin/flv_transcode.sh

It will create a toplevel “flv” directory, and put all the transcoded output in that subdirectory. Mencoder can’t directly produce FLV output, so this ffmpeg-based solution is a bit easier. Now, we just have to wait for Macromedia to support h264 files…