On March 14th, Dave wrote: I thought it was pretty clear that the purpose was to blast a hole for titles and links, and as long as I was doing that I thought I should blow a hole for categories, enclosures and anything else that might follow (like flNotOnHomePage which sounds like what Brian wants). At the time, I thought we had a breakthrough, and on the next day released Coping with Change.
Today, Dave says, I don't see the value in allowing variability, because you trade that off against complexity, too high a cost, too little gain. Sigh
I honestly don't know who designed UserTalk, but I do notice that it permits one to add parameters with default values without breaking existing code. My feeling is that if this is relevant on a local procedure call, it certainly is relevant on a remote procedure call; if not more so.
Note that the XML RPC Google Gateway is missing the ability to specify whether or not you want to filter out duplicate pages. Was that omission intentional? Will it prove to have been somethink that people will miss? All we can say at this point is that it will be hard to add in a backwards compatible way.
Thats the key point. One can design brittle interfaces and invent new ones every time something changes. Requiring your consumers to make mental leaps everytime they want to call you. Ultimately you end up with enough cruft that even Joel will agree that rewritting is the best choice.
Or you can take the daring and clever approach of using named parameter associations and requiring that the server must ignore all elements that it doesn't understand.
Note: IMHO, this is the key reason that HTML is successful today. Without that rule, it wouldn't have CSS that Mark likes so much. Or for that matter, the tables that Dave likes so much. Or even images.
9:41:37 PM 
|
