home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
tsql
/
doc
/
tsql.mail
/
000163_ahn@cbnmva.att.com _Tue Jul 6 17:05:48 1993.msg
< prev
next >
Wrap
Internet Message Format
|
1996-01-31
|
3KB
Received: from univers.CS.Arizona.EDU by optima.CS.Arizona.EDU (5.65c/15) via SMTP
id AA13145; Tue, 6 Jul 1993 21:10:33 MST
Message-Id: <199307070410.AA06086@univers.cs.arizona.edu>
Received: from att.UUCP by univers.cs.arizona.edu; Tue, 6 Jul 1993 21:10:30 MST
Subject: Re: SQL temporal language extensions
To: tsql@cs.arizona.edu
Date: Tue, 6 Jul 93 17:05:48 EDT
From: I. Ahn <ahn@cbnmva.att.com>
X-Mailer: ELM [version 2.2 PL14]
Greetings.
Please ignore it if you received this earlier.
I have some comments about the following constraints.
> * Initially, the language design will support user-defined time and
> valid time. Support for transaction time may be added later.
====================
The need for supporting transaction time is important, as numerous people
have commented in the workshop.
I believe support of transaction time will be a major contribution of the
"TSQL2 working group" in contrast with the approach of the
"SQL2/3 working group", and will serve as the basis for the
"TSQL3 working group" too.
I hope we can include this support from the beginning.
Since the semantics of transaction time is relatively simple, it may not
be difficult to agree upon a standard.
At least one commercial DBMS already supports the concept of transaction
time to some degree.
Sybase has the "timestamp" data type which is automatically updated
whenever a row containing a timestamp column is inserted or updated.
A restriction is that a table may have only one timestamp column.
What can happen is that other DBMS vendors may try to match or exceed
this feature in differing ways, and then the standard has to catch up
with the practices later on as it did for the user-defined time.
> * Discrete time will be assumed.
I hope we can discuss this further.
To assume discrete time, the standard has to specify the granularity of time.
Some of the problems with this approach are:
- Time is inherently continuous, as the real number is.
Enough said about this in the workshop.
- Granularity is an issue of representation.
- It is not consistent with the "time" data type already defined
for the user-defined time in the SQL2 standard.
The standard says that the granularity is determined by each
implementation, not by the standard, except that at least
6 fractional digits should be supported.
I am not sure we want to invent a new representation of time
in addition to the above.
In fact, SQL2 goes further and follows the convention of
the [closed, open) time.
e.g. the predicate
(time '10:00:00', time '12:00:00) overlaps
(time '11:00:00', time '13:00:00) is true
but
(time '10:00:00', time '11:00:00) overlaps
(time '11:00:00', time '13:00:00) is false.
I believe this is a good convention, and we need good reasons and
a lot of efforts to change this.
Besides, we need to address not only the query statement, but also
the data definition and data manipulation statements such as
"create", "insert", "delete", and "update".
The query statement alone will not be very useful, if there is no
support for the other statements.
I think any proposal should be evaluated in terms of how it deals
or can deal with all these requirements as a whole.
Regards,
Ilsoo Ahn (ahn@cbnmva.att.com)