Re: Sensible and NonsenSQL Aspects of the NoSQL Hoopla

Jan Hidders wrote:
> On 2013-09-04 05:28:01 +0000, Norbert_Paul said:
>>> Neither do I, and my compliments for the analyis. Am I right in reading
>>> some irritation between the lines that goes beyond a natural distaste
>>> for shoddy reasoning?
>>
>> Yes. The irriatation is based on bad personal experience within the
>> last three years. For example a (deliberate?) publication delay of
>> more than three years. There seems to be an extremely influential group
>> of researchers around the author where this kind of shoddy (thanks for
>> the new vocabulary) work is the only one accepted. Have a look at
>> "alternating hierarchical decomposition" (AHD). It just doesn't work
>> in 3D.
>
> Like this?:
>
> http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5293499&tag=1
>
> So the claim in the abstract about "scalable to any dimension" is wrong?

Yes, exactly. Bulbul makes all proofs and examples in 2D. There, indeed, all seems to work. But Take a 3D cube, say [-1,1]^3, and its center point, say (0,0). Partiton that cube into the six pyramids with each cube face as its bottom and the center point as its apex (convec closure of a face and the center). Then remove three pyramids making sure that two of them are opposite. Apply the AHD and learn that it recurses endlessly. Reason: The convex closure of the shape is equal to the convex closure of its "differece shape" (aka difference to convex closure).

An Ascii garphics illustration:

     *-------------------*
    /                   /|
   /                   / |
  /                   /  |
+-------------------+   |
|  .             .  |   |
|      .      .     |   |
|          *        |   |
|        /   \      |   +
|      /       \    |  /
|   /            \  | /
|/                 \|/
+                   +

\setlength{\unitlength}{1.5\unitlength}
\put(0,10){\line(2,1){20}}
\put(0,10){\line(5,4){25}}
\put(0,10){\line(0,1){40}}

\put(0,50){\line(5,-4){25}}

\put(30,40){\line(2,1){20}}

\put(50,10){\line(0,1){40}}
\end{picture}$$
that method may not terminate.''

to avoid backdoor arguments that sloppy specifications always leave open in the sense of "But I meant it that way ... ".

>> The problem is, when evidence is given, reviewers say that this is "too
>> abstract and technical" or they have an article wait forever to become
>> published. On the other hand, when a scientist's position is secured he
>> can make /any/ claim without any evidence. I thought this could be an
>> alternative channel here to post something critical. There is no
>> reviewing on comp.databases.theory.
>
> Indeed. But also not much of an audience. But who knows, it might come,
> and it's easier and less demanding (usually :-)) then a blog.

Google archives Usenet. Maybe one day ...

>> I will look up Date and Fabian. Are they influential?
>
> My mistake. I meant Chris Date and Fabian Pascal. In the database
> research community they have no discernable influence as far as I can tell.

Then it is not interesting to critizise them. My problem is not that someone is wrong. Then just live and let live. The problem is when seems as if (Note, "as if" doesn't say this indeed happens) a bunch of ignorants managed to establish kind of a trust.

Norbert Received on Wed Sep 04 2013 - 23:00:22 CEST