Lists Home |
Date Index |
> As for the decision, it was all over in about 45 seconds on one
> teleconference, I remember it clearly. Fairly early on in the XML
> design process. Someone (maybe me? I forget) said "er, should we do
> an API as well?" and James Clark said "isn't the idea that there's
> going to be one API that's going to work for all the different
> things you want to do kind of silly?" and everyone said "oh, right"
> and that was that.
But there's a big difference between a data model and an API onto a
data model. The Infoset, for example, isn't an API; you could imagine
many APIs onto the Infoset, which might reveal different parts of the
structure in different ways.
I guess that when I've been talking about a "data model" in this
thread, I've been talking about the Infoset-level data model rather
than the API-level data model. I fully agree that trying to define an
API-level data model would be a complete nightmare, even when you're
in a committee of 3. But agreeing on an Infoset-level data model is
"only" about as burdensome as agreeing on a syntax, at least if my
experience with LMNL is anything to go by.