Cart
[Home] [By Thread] [By Date] [Recent Entries]
From the second half of the article: "So while we're full-blown XML, what we expose through the SOAP messages and through the HailStorm data-manipulation layers has these XML boundaries coded into the system. Then, once you have this node, you can throw it into a DOM and get finer-grained access to the XML. This red and blue thing is a core piece of the HailStorm model. We used to have [other] names for these things, but the way it evolved is we would decorate our work with red and blue decoration, and we finally ended up dropping the artificial names, deciding to call it "red" and "blue" because that's how we think of it. Blue nodes have these properties and red nodes have those properties. Besides the difference in granularity, blue nodes become the cache-line boundary for replication and for taking the information offline. For that to work, the boundary-the blue node-has to be uniquely identified. So we have the "ID=" attribute-a unique ID for that blue node. We generate that ID by hand out of the XML storage system. We also have a ripple-through change number, which lets you take a cached copy of an address and look for addresses where the change number is not equal to the one that I have in my hand. I either get a new one, which will now become the definitive copy, or I get nothing and know that I'm in sync. All of the HailStorm information follows the same basic pattern. It is designed to be taken offline. It's designed to be highly cacheable so you can do things like take your address book into your cell phone, update it from there, and keep up to date with the master." And they reinvented the geo-locator nodes of public safety systems. Interesting. Comment to MS: public safety databases already have those mappable geo-locators. We thank you for your adulation and support. :-) len
|
