Re: The Practical Benefits of the Relational Model
Date: 19 Oct 2002 03:36:14 +0200
Message-ID: <3db0b70e_at_news.uia.ac.be>
In article <aopeod$1c5i$1_at_sp15at20.hursley.ibm.com>,
Paul Vernon <paul.vernon_at_ukk.ibmm.comm> wrote:
>"Jan.Hidders" <hidders_at_hcoss.uia.ac.be> wrote in message
>news:3dafeecb$1_at_news.uia.ac.be...
>> >> >At the cost of other users having to wait.
>> >>
>> >> No, the *illusion* that the other users have to wait. How much the
>> >> other users really have to wait depends upon the locking protocol.
>> >
>> >And on how trustworthy the other users are.
>>
>> Not necessarily as much as you seem to claim. How much you depend on that
>> depends upon your locking protocol, and for certain optimistic locking
>> protocols this is just as much the case as it is in your proposal.
>
>Could you point me in the direction of such protocols?
See for example 'timestamp-based concurrency control' in Ullman's Principles.
>> So again, this not an intrinsic problem in the concept of user visible
>> transaction. Moreover, such optimistic protocols are not under all
>> circumstances ideal and in some sense your proposal means that this is
>> fixed. So the concept of user visible transaction is in that respect
>> better because it leaves this choice open
>
> open to abuse by untrustworthy users
You're always open to such abuse, even in your proposal. With optimistic locking a user can sabotage other transactions by causing a lot of conflicts. The only way to avoid that is by not letting the database guarantee anything about whether the values you just read are still valid or not when you commit your writes.
- Jan Hidders