From: JB
To: Hugh Darwen
CC: Editor
Date: 1 Apr 2004
Thank you both for your attention and the time spent
responding to my inquiries. It will take me some time to absorb all the
information below.
I'm reading Mr. Date's INTRODUCTION TO DATABASE
SYSTEMS, 7th Ed. Unfortunately, I obtained it just before the 8th
edition became available. I skipped chapter 22 on temporal data, assuming that
the information is superseded by TEMPORAL DATA AND THE
RELATIONAL MODEL. Should I read the chapter even though I have already
read the other book?
I don't reject the term "type" in favor of the term
"domain." I just want the terms I'm using be precise even out of
context.
I have been thinking about the concept of a context free
language for quite some time. Let there be a context free language. Let each
user define a set of transformations from the context free language into a
dialect for that user. Allow any user to create new words in her dialect that
automatically define unique words in the context free language. Allow any other
users to manually discover newly created words in the context free language and
to define new transformations for those words into his dialect. I'm sorry I
can't explain this idea more clearly. I've only just begun to work on it. I
have yet to read any good texts on programming language design, but I will
begin reading the references suggested in THE THIRD MANIFESTO
as soon as possible. Can you recommend any references for programming language
design?
I asked my original questions below because I was considering
implementing a DBMS. I've since seen that there are several projects attempting
to build either DBMS kits or DBMSs based on The Third Manifesto. I've begun
writing my own implementation, so I hope one more attempt won't hurt! As a
result, I've given up on the idea of infinite domains and relations, mostly for
practical reasons. I will return to that area of study at a later date. I would
love to hear any comments you have or pointers to other research about the
subject.
I understand that TRUE and FALSE are the basic concepts. Can
TABLE_DEE and TABLE_DUM then be considered possible representations of TRUE and
FALSE respectively?
Your comments have helped me clarify my ideas.
From: Hugh Darwen
To: JB
Cc: Editor
Date: 4 Apr 2004
Yes, Chapter 22 is definitely superseded (spelled thus,
according to my dictionaries) by the book.
It is also superseded by the same chapter in the 8th edition, as Date
told me he had to make many corrections as a result of our work on the book.
Okay, but I'm intrigued to know why you think
"domain" more precise than "type".
Sorry, but I don't understand your idea yet. In answer to your question, see reference 9
in Appendix K.
No, one more attempt won't hurt. Nor will 100 more as far as I'm concerned, so long as they are
all Ds! Please keep me posted,
especially when you would like us to place a reference to your project at our
web site.
Because types are used in the definition of a possrep and a
possrep is needed in the definition of a type, we have to take something as
just given to get off the ground. I can
see how starting with TRUE and FALSE works, because if BOOLEAN is just given
(with no THE_ operators) we can make it a possrep for BIT.
TABLE _DEE and TABLE_DUM seem to need our
concept of the relation-type generator.
I feel a bit uneasy about relying on the concept of a type generator
before I have any types at all! I would
be interested to hear your reaction to that thought.
Your questions help me to clarify mine, too!
Posted:
05/28/01
