Re: Splitting up the spec

Julian Reschke wrote:
> Another issue of course is IE's behavior with respect to stripping 
> "ignorable" whitespace (IE ignores more than it should).
> 
> So yes, all these are instances of under-specification, but not 
> *necessarily* caused by spec fragmentation.

While this is true, here is another statement that seems to me to be on 
the same level of truth:

   While Bob was hit by a trunk and killed, death is not *necessarily*
   caused by being hit by a truck.

This statement is, of course, true, for all possible meanings of 
"necessarily".  It's possible to die without being hit by a truck.  It's 
possible to be hit by a truck and not die.

But the fact is, the conditional probability of Bob being dead given 
that he's been hit by a truck today is much higher than the a priori 
probability that he was dead that we'd assign before we heard the truck 
news.

Similarly, the probability of things being under-specified goes up, 
drastically, once multiple specifications are interacting.  Based on 
prior observations, I'd assign a probability well north of 0.5 to such 
under-specification occurring at any junction of two independent (really 
independent, not completely tied to each other and modified in lockstesp 
while being edited by the same Working Group and editor in two separate 
documents) specifications.  It's a gut feeling number, but I honestly 
can't think of _any_ cases where this happens that are well-specified, 
much less remain well-specified as the specifications involved evolve.

-Boris

Received on Tuesday, 25 November 2008 01:16:03 UTC