The basic idea of encryption is this- you have a message, and you want to protect it. So, what you do is generate noise, or something that looks incredibly like noise, and according to some mathematical process, combine your message with the noise, hoping that the output looks indistinguishable from noise.
The most successful version of this is the One-Time Pad encryption. Sally creates a One-Time Pad. This is done by taking random values (like drawing bingo balls or cards from a deck) and making a long list of them, converted to letters. This pattern is noise. She then gives a copy of this one-time pad to Bob.
When Sally wants to send a message to Bob, she writes her message, then takes the first letter in the pad, adds it to the first letter in her message, and so-forth until she's combined every letter in her message with every letter in the one-time-pad.
This method of encryption is unbreakable. There is no way to attack it based on letter frequencies- the only way to do so is to steal a copy of the pad. Unfortunately, each recipient must have a copy of the pad, and further, must dispose of the pad each time a message is encrypted, thus creating the need for large numbers of these pads to be generated.
I came up with an interesting revision to this idea, using your friend and mind, RSS.
The Overtime Pad works like so. Bob and Sally agree on a set of URLs to RSS newsfeeds. This will be the "key" for their encryption (giving us a keyspace equal to all permutations of all newsfeeds on the web). When Sally encrypts her message, she goes to each newsfeed and takes the most recent post. She takes each one of those posts, and combines them somehow, along with other variables, like the length of the post, the URL itself, and so forth- until she gets the most random One-Time-Pad possible.
She encrypts it, using the One-Time-Pad scenario, dates the result, and sends it to Bob. Bob can then open the message, not the time it was sent, and retrieve the nearest post to that time (without going later), use the same mathematical mumbo-jumbo and create his own copy of the One-Time-Pad and voila! He can now decrypt the message.
Perks- if you can create a random enough pad from these multiple data sources (which is a tricky proposition, but I have a handy little program called Ent to guide me towards the most noisy results from these newsfeeds), you have a well-nigh unbreakable code, except for a completely unreasonable brute force attack involving reading every RSS feed on the web, which is very likely impossible- especially if a few sites start exporting random data as RSS feeds, which would become a useful service with this sort of encryption available.
This method puts an automatic window on when these messages can be decrypted (plus or minus, depending on your needs). Most RSS feeds will expire posts older than the last 25 or so, which means you have from the time the source post was made, to when it drops off the feed to read the message.
This is a good system for rapidly exchanged email, but not for archiving, or anything like that.