home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
tsql
/
doc
/
tsql.mail
/
000150_atuzhili@squar….stern.nyu.edu _Mon Jun 7 15:58:14 1993.msg
< prev
next >
Wrap
Internet Message Format
|
1996-01-31
|
3KB
Received: from SQUARE1.STERN.NYU.EDU by optima.CS.Arizona.EDU (5.65c/15) via SMTP
id AA16231; Mon, 7 Jun 1993 12:55:47 MST
Received: by square1.stern.nyu.edu (4.1/1.34)
id AA21417; Mon, 7 Jun 93 15:58:14 EDT
Date: Mon, 7 Jun 93 15:58:14 EDT
From: atuzhili@square1.stern.nyu.edu (Alexander Tuzhilin)
Message-Id: <9306071958.AA21417@square1.stern.nyu.edu>
To: tsql@cs.arizona.edu
Subject: Re: TSQL Benchmark STATUS
The latest messages from Christian and Rick concerning the TSQL
Benchmark Queries have justifiably asked those who have not yet
contributed a set of queries to the benchmark to please do so by the
indicated deadline. We sat down together this morning to do just
that, and in trying to fully understand the categories we got to
reflecting again upon the purpose of the benchmark as a whole.
Basically we ran up against the same discomfort with this benchmarking
effort that we indicated in an earlier message in this forum. We
would like to restate our reasons for this discomfort.
We do not understand the claim in the Draft document that the
benchmark will contribute to evaluating the "user-friendliness" of
temporal query language proposals. User-friendliness is not
typically measured in this way.
It still seems to us that the whole idea behind this "semantic
benchmark" is to provide some measure of what should be expressible in
a temporal query language. As we indicated in our earlier message, we
believe that a list of queries is not the appropriate way to provide
such a measure. One of the criteria mentioned for the taxonomy (p.5)
is that it should provide "comprehensive coverage of benchmark
queries." We do not feel that a classification that does not have
some sound theoretical basis can be considered comprehensive. For
example, in the discussion of the Valid-Time Selection taxonomy (p.7,
and Figure 4), "comparison operators" are mentioned, yet only a few
examples (spanExceeds, preceded, meets, etc.) are provided. What is
the complete set of these operators? In temporal logics there are
dozens of different operators that have been defined (since, until,
before, after, while, always, sometimes, etc.), and different subsets
of these operators, combined with different underlying models of time,
have provably different expressive powers. Without some clear
theoretical guidance, there is no basis for the selection of the 30
categories described in the benchmark document, nor any basis for
formulating "five example queries" in any meaningful way.
We don't want to sound too negative in this message, but we would like
to encourage everyone to think again about the purpose of this
benchmark, the issue of the "comprehensiveness" of the benchmark, and
the issue of the expressive power of temporal query languages. We hope this
message will be constructive, and will generate discussion and comments.
We welcome both.
Alex Tuzhilin and Jim Clifford