Progress on the API front
After a lot of work and testing, I think I finally have the MetaWebLog API working on the production server. I've tested with both ecto and BlogJet - BlogJet now works against my server, fetching posts with categories. ecto fetches posts and categories, but doesn't match them up. I've been working with folks from ecto and BlogJet - they have both been very responsive and helpful. The problem is this sorry excuse for a spec from Dave Winer. He's periodically said he doesn't feel like he gets enough credit for this stuff; if I were responsible for that crappy API, I'd stay very, very quiet. Every publisher and blog server vendor has had to make guesses about how it works, because the "spec" is pathetic. In my server, I'm returning categories under three different names now - a guess based on the spec, one way for ecto, and another way for BlogJet. The only saving grace is that clients expect a dictionary, so they pretty much ignore data they aren't looking for.
None of this should be necessary...
Comments
Architecture specs
[ Rich Demers] March 8, 2005 11:48:10.000
Comment by Rich Demers
If anyone is interested in a successful way to specify a complex communications architecture, take a look at this. It describes work I did throughout the 1980's to enable mainframe systems to easily access each other's files and databases. Much of it is now of little interest (record oriented files?), and it used pre-Internet communications protocols (SNA LU 6.2), but what's worth looking at is the method I used to document the architecture.
In brief, I borrowed ideas from Smalltalk. The entire architecture was specified by a dictionary consisting of menu objects, help objects, and class description objects that specified the clients, servers, messages and protocols of architecture. It used polymorphism and inheritence to keep things simple and consistent, and the on-line version was tightly hyperlinked.
Did it work? You bet it did -- even though I didn't have a running Smalltalk system available to me at the time. Roughly 40 different products were built and they were able to work together without too many problems. I believe some of them are still in use.
But it's probably expecting too much to think anyone in our industry would ever look at anything done in the past. Sigh.
[] January 1, 1901 0:00:00.000