From: Clark C . E. <cc...@cl...> - 2001-11-28 05:26:34
|
Oren has brought up a serious point that we have to seriously consider. At the information model level, it would be nice to be able to assert that two scalars are equal if their strings are. Obviously this won't work... so we need to further complicate the transfer method by extending each one with a an equivalence relation requirement... given any two string representations, it must provide for a way to see if those two strings are equivalent. Quite clearly we know that two strings are not equivalent when they have a differnt string representation. I don't like this, but I don't see any alternative. Unfortunately, it means that there is *no way* to ensure that a mapping is valid if it has keys that are not part of a registered transfer method... Hmm. I guess we can use a fall back in this case, and just do a string comparison. .... That said... it's complicated enough when scalars stay scalars. Now, if we have scalars turning into collections... that's a horse of a new level of complexity.... and I'm not sure that we are prepared to handle it at this point. Therefore, I'm asking that we table all of our cool collection short-hands till the next version. .... Also, I'm proposing that "quoted" scalars be added back in to the syntax productions. It's clear that we need them, and they are *way* too magical to be kept as a transfer method. Best, Clark |