home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
tsql
/
doc
/
tsql.mail
/
000082_nls@cse.iitb.ernet.in _Wed Apr 14 23:59:01 1993.msg
< prev
next >
Wrap
Internet Message Format
|
1996-01-31
|
4KB
Received: from relay1.UU.NET by optima.cs.arizona.edu (5.65c/15) via SMTP
id AA25586; Wed, 14 Apr 1993 23:59:01 MST
Received: from spool.uu.net (via localhost.UU.NET) by relay1.UU.NET with SMTP
(5.61/UUNET-internet-primary) id AA02803; Thu, 15 Apr 93 02:58:45 -0400
Received: from sangam.UUCP by spool.uu.net with UUCP/RMAIL
(queueing-rmail) id 012324.1632; Wed, 14 Apr 1993 01:23:24 EDT
Received: by sangam.ncst.ernet.in (4.1/SMI-4.1-MHS-7.0)
id AA23578; Wed, 14 Apr 93 10:07:26+0530
Received: from kailash.cse.iitb.ernet.in by iitb.ernet.in
SENDMAIL Version (4.1/SMI-4.1-MHS-7.0)
id AA13672; Wed, 14 Apr 93 10:09:19+0530
Received: by kailash.cse.iitb.ernet.in (4.1/SMI-4.1)
id AA13423; Wed, 14 Apr 93 10:10:38 IST
Date: Wed, 14 Apr 93 10:10:38 IST
From: nls@cse.iitb.ernet.in (N L Sarda)
Message-Id: <9304140440.AA13423@kailash.cse.iitb.ernet.in>
To: "Christian S. Jensen" <info-tsql-sender@cs.arizona.edu>,
tsql@cs.arizona.edu
Subject: Re: TSQL Benchmark, Task 3.
The proposed taxonomy of the benchmark queries is a well
thought-out and exhaustive work. Being designed for TSQL
bechmarking (ie., benchmarking of SQL-based temporal language
proposals), it seems natural to derive the taxonomy from
SQL syntax. However, I think the problem can be approached
from another angle, and I hope others also consider it
worthwhile to explore.
Basing the benchmark queries on SQL syntax leads to bias towards
1NF relations. It also leads to fairly large categories of queries,
many of which would naturally be supported in any extension
to SQL. We can assume most extensions to SQL to permit partial
and complete list of attributes to appear in output, referencing
valid time attributes both in output and in predicates, and
capability for cartesian product. Many categories will be easily
expressible in most extensions to SQL. How well they provide for
time domain operations and temporal algebra level operations
(such as time-slicing, coalescing, temporal natural join, etc.)
should be our focus in benchmarking.
We could view the problem from user requirements point of view
and consider interesting set of query categories with
emphasis on temporal dimension. Some exercise like this may
allow us to study how complete is the proposed categorization
(20 categories based on projection component X 27 categories
based on selection component).
The following few categories occur to me from considerations
of end-user requirements :
. temporal predicates for temporal selection : operations
on valid, user-defined times or their mix
. temporal projection with / without coalescing
. time-slicing with time in output contained in time-slice
interval
. retrieving current or past data
. retrievals based on discontinuities in data (eg., retrieve
something when John was not in Toy dept)
. concurrent (natural temporal) join based queries
. retrieving data in an interval based on condition which must
hold at any, none or all instants of another interval
. .....
The categorization above is obviously not orthogonal (they are
not even in order of complexity). They appear to be useful type
of queries and we would like to support them in a simple and natural
fashion in a temporal-SQL.
What is suggested above does have its limitations (eg.,
exhaustiveness and orthogonal categorization), but appears to
merit attention.
Comments are welcome.
Nandlal Sarda
(nls@cse.iitb.ernet.in)
--- April 14, 1993 ---