Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Motion P1788/M0032.01:MidpointMeaning -- discussion period begins



> I would accept a change that it is limited to
> bounded non-empty intervals at level 1.

Thank you. I agree entirely with the modified motion.

  -Dima

----- Исходное сообщение -----
От: intervals08@xxxxxxxxxxxxxx
Кому: dmitry.nadezhin@xxxxxxxxxx
Копия: stds-1788@xxxxxxxxxxxxxxxxx, intervals08@xxxxxxxxxxxxxx
Отправленные: Суббота, 10 Март 2012 г 1:01:56 GMT +04:00 Абу-Даби, Маскат
Тема: Re: Motion P1788/M0032.01:MidpointMeaning -- discussion period begins

> Date: Fri, 9 Mar 2012 05:47:16 -0800 (PST)
> From: Dmitry Nadezhin <dmitry.nadezhin@xxxxxxxxxx>
> To: <intervals08@xxxxxxxxxxxxxx>
> Cc: <stds-1788@xxxxxxxxxxxxxxxxx>
> Subject: Re: Motion P1788/M0032.01:MidpointMeaning -- discussion period begins
> 
> Dan,
> 
> I agree with your Level 2 definition of mid .
> 
> And I have a small note on level 1 definition.
> John's Level 1 draft defines domain level 1 "mid" function as
> nonempty bounded intervals only.
> This motion defines it for semi-infinite intervals also.

	This is so & I was hoping no one would notice. :-)

> 
> This means that Level 1 mid(X)=mid([u,+oo[)=(u + oo)/2 = +oo .
> This seems natural.
> However, it violates the formulated property
> mid(X) \element-of (X1 \intersect X2)
> +oo can't belong to (X1 \intersect X2) because level 1 intervals are
> sets of reals.
> 
> Is the domain of Level 1 definition of "mid" in the scope of this motion ?

	It is & I would accept a change that it is limited to
	bounded non-empty intervals at level 1.

> 
>   -Dima
> 

	Preamble to all: In order to address Dmitry's comment
	on its merits, I am about to discuss some controversial
	stuff that has absolutely NOTHING to do with the motion
	in question.  Please do not start a long discussion in
	this forum.  If you must, talk amongst yourselves.


	Dmitry,

	It DOES violate the notion of semi-infinite sets that
	live at level 1.  At least, those that live within IR.
	And if you would like me to modify the proposal to
	include only bounded non-empty intervals within IR at
	level 1, I would accept that as friendly.  They can
	be reintroduced at level 2 in much the same way as
	midpoint(Entire) was done.

	Frankly I did it intentionally hoping no one would
	notice.  You see, I have been struggling with issues
	about level 1 that are in the relm of analysis & limits
	& other things about how we treat infinity.  I have
	never wanted to discuss them because I feel they are
	largely irrelevant to our standards work & I did not
	want to start yet another lengthy philosophical
	argument with no good resolution or utility to our
	work.

	Still, now that you have asked, level 1 also has
	intervals that live in IRbar.  And, strictly speaking,
	that would mean sets that include infinity as an element.
	But we have chosen to take the interpretation that a
	semi-infinite interval is open at the far end such that
	infinity is NOT an element of that set.  Fine & good.
	I can live with that.

	However, we still have some funny sets that live at
	level 1.  Within IR we have sets like bigun(r) = [r,2r].
	That means, for any finite set, X, you can name, there
	is an s > sup(X) such that bigun(r) > X for all r > s.
	Then within IRbar, the set bigend = limit (r-->inf) bigun(r)
	can be defined.  It is a set that contains no Real
	elements yet it is different than the empty set in that
	it is known to be larger than any finite set you can
	name.

	A funny set.  At least to me.

	One could also consider the implications of semi-infinite
	sets like bigsemi(r) = [r,+inf) with slightly different
	lines of reasoning.

	Anyway, rather than introduce these notions to the
	pointless controversy I KNOW they would engender, I
	fudged on the definition.

	I can go back if you think its important.

	For Juergen & Baker, the changes would be to limit the
	scope of the level 1 midpoint to bounded non-empty
	intervals.  And then to introduce the semi-infinites &
	Entire at level 2 as projections onto some finite
	explicit interval system IF over some floating-point
	system F.

	But we need NOT have a big argument about it.

	Please.

				Dan