home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
tsql
/
doc
/
tsql.mail
/
000124_gadia@cs.iastate.edu _Thu May 13 20:04:09 1993.msg
< prev
next >
Wrap
Text File
|
1996-01-31
|
3KB
|
80 lines
Message-Id: <199305140104.AA28613@optima.cs.arizona.edu>
Received: from ren.cs.iastate.edu by optima.cs.arizona.edu (5.65c/15) via SMTP
id AA28613; Thu, 13 May 1993 18:04:09 MST
Received: by ren.cs.iastate.edu
(16.8/16.2) id AA15862; Thu, 13 May 93 20:04:10 -0500
From: Shashi K. Gadia <gadia@cs.iastate.edu>
Subject: Benchmark/"Taxonomy"
To: tsql@cs.arizona.edu
Date: Thu, 13 May 93 20:04:09 CDT
Cc: gadia@cs.iastate.edu
Mailer: Elm [revision: 70.30]
I see that a lot of effort has gone
in to "taxonomy". Unfortunately
there is very little here that I agree
with.
A. It is very narrowly focused, and
disregards the notion of orthogonality.
If there are 4320 different categories,
I feel there is something fundamentally
wrong with the approach.
B. I find the option "imposed" in Figure 2
quite questionable. It is to be taken
with a grain of salt. This amounts to
cooking new information rather than
retriving information existing in the
database. I can not see how we can elevate
it to the same level as other options.
C. The option "none" under the valid-time
component raises an important question.
This allows timestamps to be erased from
a relation. So does it mean we have two
types of relations: temporal relations
and "non-temporal" relations? If these
non-temporal relations can be computed,
can they be mixed with temporal relations
to answer additional queries? If no,
then the non-temporal relations are only
terminal, and I have no problem with this.
But if they are not terminal, then you
could retrieve misleading information.
Its better to be conservative and simply
not allow nonterminal-nontemporal
relations. Otherwise every participant
should take an clear stand over this
issue.
D. Although the remarks in this subsection
concern our approach, but it also relates
to most other approaches, perhaps in different
measures.
If I disregard the following options:
imposed from Figure 2,
ordering-based from Figure 4,
interval from Figure 4,
then in our TempSQL there is only 1 and
NOT 4320 different categories, all handled
in a uniform manner. This leaves me wondering
that if I want to propose a query, why should
I have to bother about associating one of the
4320 categories!
Summary.
--------
I appreciate a need for a structure. However,
I do not know what that structure is, but
if there is one, I strongly feel that it will
come from English (a natural language).
The bottom line is that if someone
suggests a new query form(s), we are obliged
to include it in the benchmark, irrespective of
whether we have understood the underlying
classification.