home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
tsql
/
doc
/
tsql.mail
/
000059_gadia@cs.iastate.edu _Wed Mar 31 19:13:28 1993.msg
< prev
next >
Wrap
Text File
|
1996-01-31
|
5KB
|
151 lines
Message-Id: <199304010142.AA02832@optima.cs.arizona.edu>
Received: from ren.cs.iastate.edu by optima.cs.arizona.edu (5.65c/15) via SMTP
id AA02832; Wed, 31 Mar 1993 18:42:27 MST
Received: by ren.cs.iastate.edu
(16.8/16.2) id AA07427; Wed, 31 Mar 93 19:13:28 -0600
From: Shashi K. Gadia <gadia@cs.iastate.edu>
Subject: Benchmark/Our remarks
To: csj@iesd.auc.dk (Christian S. Jensen)
Date: Wed, 31 Mar 93 19:13:28 CST
Cc: tsql@cs.arizona.edu
In-Reply-To: <199303291426.AA22405@iesd.auc.dk>; from "Christian S. Jensen" at Mar 29, 93 4:26 pm
Mailer: Elm [revision: 70.30]
Dear Christian:
Our comments on your most recent (March 29, 1993)
proposal for Benchmark document are as follows:
1. I have decided to withdraw the benchmark effort
which was initiated originally by Sunil Nair
and me. I have already informed Abdullah Tansel
and Gene Wuu of this decision, as they were
participating in our original effort. I have
recommended to them that they should directly
participate in the current effort. I will
also be informing Sarda of this, as he was
also participating on our effort. Having two
efforts with similar scopes makes no sense.
From now on you will have our full participation.
2. I would like Sunil Nair to be included as a coauthor,
as this effort was originated by the two of us.
3. Please include the following in the tsql alias:
Sunil Nair: snair@cs.iastate.edu
Nandlal Sarda: nls@cse.iitb.ernet.in
Lorentzos: I do not have his email handy.
Our (Sunil Nair and Mine) comments are as follows:
(These comments are only for the current version
being proposed.)
Contents of relations.
----------------------
It would be difficult to state the contents of relations
in a natural language, and even more difficult to read.
Such information has been listed in tables even before
(electronic) databases were invented. If this alternative
is not acceptable to you, we suggest that the contents
be repeated in a tabular form.
Transaction time queries.
-------------------------
It depends upon what kind of queries you
are trying to exclude. Our (Bhargava-Gadia)
VLDB'89 (also Feb 93 TKDE) shows that much of transaction
time database can be queried in a manner identical to
valid time. In fact every valid-time query can be asked
for transaction time. For example, the
valid time query "when did John work in Toys" has the
corresponding transaction time counterpart "when was John believed
to have worked in Toys". In our languages the two queries are truly
orthogonal, as the sql-query for the two are identical.
We would like the document to be more clear about
what specific examples of queries are being excluded.
You may also recall our discussion on Glossary when
we were discussing rollback database v/s transaction
time database, and decided in favor of transaction time
database.
Recursive queries and aggregates.
---------------------------------
We do not have strong feelings about excluding these
two forms of querying. But please note that the queries
are English based, and those queries which someone
can express without using recursion/aggregation
should be allowed in the first draft.
Nested queries.
---------------
Nesting of queries is heart of relational
systems. Same English query expressed in
different ways may use different levels of
nesting. This concept is central to user
friendliness of relational systems.
Staying within English, this is an
artificial requirement. We believe even
SQL2 allows considerable nesting. We know that
SQL3 is set to allow full fledged nesting.
Each relation is used only once in a query.
------------------------------------------
This is also artificial, and rather restrictive.
The above comments under "Nested queries" are
also applicable here.
Relation/attribute names.
-------------------------
We suggest that we require relation names to
be started with a lowercase letter, and
an attribute name started with an uppercase
letter. This makes queries easier to read.
Example: mgr is a relation, and Manager is
an attribute. This is only an informal
convention.
Time attributes/ time alignment.
---------------------------------
We like the suggestion that user-defined time
value may be compared with valid time. It
would help to emphasize this fact, by perhaps
saying that the two times originate from
the same source. We would like to see an
attribute which has a temporal element as its
value. An example would be "Leave" attribute,
allowing for a query such as "when John
was on leave ...".
mgr relation.
-------------
We would like the functional dependency
Manager --> Dept to be added to the mgr
relation. This is an important issue as
it would allow people the option of taking
a stand on homogeneity and role of keys in
temporal relations.
3rd Normal form.
----------------
As in Snodgrass' recent message, we feel
that emp should be in 3nf.
Regards
Sunil Nair and Shashi K. Gadia