Re: DL 909

On Wed, 2003-08-27 at 10:14, Jeremy Carroll wrote:
> I agree with Sean that DL 909 is currentl flawed.
> I prefer a small change over making it extra credit.
> 
> In particular I am pretty sure that the number which is currently 10 
> billion was intended to be 1 billion (i.e. fitting within 31 bits).
> 
> I am happy to make this even smaller if for example WG members felt that 
> more than 23 bits was unreasonable.
> 
> This number is used in the test to distinguish finite from infinite.
> i.e. a set with cardinality of less than this number is known to be finite. 
> However the number is sufficiently high that an exhaustive analysis of each 
> possibility is unlikely to be practical.
> 
> ===
> 
> We may at a later stage in the CR discover that no implementations are 
> implementing such a distinction, hence we may choose to make this extra 
> credit anyway, but for the more compelling reason that the heart of the 
> test is hard, rather than the superficial reason that the test overflows a 
> register somewhere.
> 
> Propose - modify test dl/909 to use 1 billion instead of 10 billion.

Sounds reasonable to me.

I'm not sure who this proposal is made to.

The thing is currently a proposed test, not an approved test, yes?

As editor, if you think a proposed test should be changed,
change it. If you're soliciting advice first, very well.
And do keep us notified of changes, be prepared to
negotiate them, and all that. But don't put the WG
in the critical path for every editorial change, please.

If you're proposing that the WG approve the test as
ammended, very well, but please be a bit more clear.


> Jeremy
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Wednesday, 27 August 2003 11:56:12 UTC