The arrival and insanely-fast growth of syndication/RSS technology brings a New Thing to the Internet. Until recently, there was only one messaging architecture known to work at Internet scale. That was store-and-forward, as in how email works: I send my message to a computer near me, which stores it and sends it to another computer near you (with retries and so on as appropriate), which stores it, and you retrieve it from that computer. Syndication has proven that a different model, post-and-poll, scales up too. It works like this: I post some data which contains either messages or message pointers to a public place, and you poll periodically to see what’s new. The difference is that store-and-forward supports anybody-to-anybody message traffic, while post-and-poll assumes separate communities of senders and receivers. People who are designing message interchange frameworks that might need to become Internet-scale should consider this, and be careful of architectures that don’t fall into one of these two baskets, because nothing else has yet been shown to work. [Update: Somehow I forgot to credit Mark Hapner of Sun who pointed all this out to me, presenting it as something obvious, not a discovery; but then I realized that it hadn’t been obvious till he did so.]
