home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
tsql
/
doc
/
tsql.mail
/
000050_jcliffor@is-4.stern.nyu.edu _Thu Mar 25 15:18:20 1993.msg
< prev
next >
Wrap
Internet Message Format
|
1996-01-31
|
4KB
Received: from IS-4.STERN.NYU.EDU by optima.cs.arizona.edu (5.65c/15) via SMTP
id AA07356; Thu, 25 Mar 1993 13:11:04 MST
Received: by is-4.stern.nyu.edu (4.1/1.34)
id AA24453; Thu, 25 Mar 93 15:18:23 EST
Date: Thu, 25 Mar 93 15:18:20 EST
From: Jim Clifford <jcliffor@is-4.stern.nyu.edu>
To: tsql@cs.arizona.edu
Subject: Benchmark Schema and Queries
Message-Id: <CMM.0.90.2.733090700.jcliffor@is-4.stern.nyu.edu>
I have been thinking some more about some of the issues that have been
raised in this forum with respect to the development of a benchmark schema
and set of queries. I thought it time to post.
First, Kalua and Robertson's message expressed many sentiments that I share.
Basically, I do not feel that there is much to be gained from leaving
out of the schema -- in the name of a "phased approach" -- semantic
aspects which seem to be universally regarded as important in a
temporal database. However, I also agree with Rick Snodgrass' comment
that we do not need to have too much "functional redundancy."
It seems to me that a good way to proceed would be to first
try to come up with the "functional" (or "semantic") aspects that we
want (and can agree) to include, and then come up with an example
schema, rather than starting with the schema.
Here is a list of what I gather to be the aspects on the table:
(1) single-valued, time-varying attributes that change
discretely over time, e.g., SALARY
(2) single-valued, time-varying attributes that change
continuously over time, e.g., TEMPERATURE
(3) single-valued, time-invariant attributes, e.g. BLOODTYPE (instead
of GENDER), as suggested in (my half of) Clifford, J. and Tansel, A.
``Towards an Algebra of Historical Relational Databases,'' {\em Proc.
ACM SIGMOD}, Austin, 1985, 247-265.
(4) single-valued, so-called (I still hate this term) "user-defined"
temporal attribute, e.g. BIRTHDATE
(5) multivalued, time-varying attributes: e.g., SKILLS
***************************************************************************
Some of the other issues raised seem to me to relate to the features of
the QUERY LANGUAGE, rather than the SCHEMA:
(a) aggregates -- I see no reason not to include examples of these queries
(b) multiple references to the same relation in a query -- excluding
this seems completely artificial to me.
(c) queries that respect/do not respect the temporal integrity of
historical values (grouped -vs- ungrouped) -- again, I see no reason
to exclude these.
I guess overall my feeling about the query language is that just about
ANYTHING ought to be allowed,and that when we classify the queries we
will indicate which "semantic features" a particular query is
presupposing are incorporated into the model. Then, if the model has
those features, the query can successfully be answered, otherwise it
may be only partially successfully answered.
***************************************************************************
SOME SEEMINGLY AGREED-UPON ISSUES WITH REGARD TO THE SCHEMA for PHASE I
(1) Should not include transaction time
(2) Should not include schema-versioning
***************************************************************************
SOME SEEMINGLY OPEN QUESTIONS WITH REGARD TO THE SCHEMA
(I) Is there an example of a multivalued "user-defined" temporal
attribute?
I don't know.
(II) Should we include a greater variety of time-varying attributes, since it
is well known that the possible relationships between data and time are many.
For example, SALARY (above), is itself an aggregated value (generally over a year),
TEMPERATURE is not. Other types of values are aggregated over different granularities
(e.g., DAILY-SALES VOLUME), etc.
I think so, because I think that this is an area that is still in need of
greater understanding.
***************************************************************************
I welcome any comments.
--jim clifford
************************************************************************
Jim Clifford jclifford@stern.nyu.edu
Associate Professor TEL: (212) 998-0803
Department of Information Systems FAX: (212) 995-4228
Leonard N. Stern School of Business
New York University
Management Education Center
44 West 4th Street, Suite 9-74
New York, NY 10012-1126
************************************************************************