Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Row locking and serializability

Re: Row locking and serializability

From: David Cressey <dcressey_at_valinet.com>
Date: Sat, 26 Jun 1999 08:10:40 -0400
Message-ID: <xu3d3.91653$_m4.822744@news2.giganews.com>


Iggy,

Thanks for posting the excerpt from the SQL-92 paper.

It would appear from the excerpt as though there were no greater level of serializability possible than the one given in the definition in the SQL-92
standard.

Now the question becomes whether "transaction isolation" and "serializability"
refer to the same thing. I think they do not. Obviously, transactions that are
not serializable with respect to each other may not be considered isolated from each other.

But the question remains whether a transaction must necessarily block another
transaction in order to guarantee serializability.

When I first read "readers do not block writers" I read that as meaning "read only transactions do not block read write transactions" . I didn't realize
that others would read it as "reads in the context of a read write transaction
do not block writers".

Reads executed in the context of a read write transaction must lock writers. Reads executed in the context of a read only transaction need not block writers,
provided that the read only transaction may be given a consistent view of the database, by retrieving the value of any data item read as of the point in time when the read only transaction started.

This is what is done in the case of what DEC RDB called a "snapshot transaction"
DEC RDB version 1 (sometime around 1982) had snapshot transactions.

There are other issues that have been raised in this thread. I may get to some of them
in a different message.

(In 1994, DEC sold DEC RDB to Oracle, which renamed it "Oracle RDB". I am sorry
if this causes any confusion with Oracle RDBMS, which most people know as "Oracle")

Regards,

    Dave

iggy_fernandez_at_my-deja.com wrote in message <7k37ol$rp7$1_at_nnrp1.deja.com>...
>RE: Definition of Serializability in Subclause 4.28 of the SQL92 standard.
>
>Here's what I have. I am referring to the copy of the SQL92 standard that I
>found at the address
>"http://sunsite.doc.ic.ac.uk/pub/computing/programming/languages/perl/db/re
fi

<snip> Received on Sat Jun 26 1999 - 07:10:40 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US