home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.databases:6171 comp.databases.theory:370
- Newsgroups: comp.databases,comp.databases.theory
- Path: sparky!uunet!decwrl!world!edwards
- From: edwards@world.std.com (Jonathan Edwards)
- Subject: Re: Relational Queues
- Message-ID: <Bt6K3v.GrA@world.std.com>
- Organization: The World Public Access UNIX, Brookline, MA
- References: <1992Aug13.195729.19884@oracle.us.oracle.com> <1992Aug14.202406.17593@mp.cs.niu.edu> <1992Aug16.062239.24049@oracle.us.oracle.com>
- Date: Tue, 18 Aug 1992 12:59:55 GMT
- Lines: 47
-
- In article <1992Aug16.062239.24049@oracle.us.oracle.com> mfriedma@uucp (Michael Friedman) writes:
- [proposal to handle FIFO queues, particularly the dequeue operation, by
- putting various tag columns into a relation]
-
- Let me complicate the problem, by requiring TRANSACTION CONTROL of all
- queue operations. Pretend that these queue entries are your paycheck being
- direct-deposited [thats actually what my systems are used for]. So you want
- to be able to dequeue from one queue and process the entry all as part of a
- standard database transaction. Transactions eliminate the ability to use tags
- to declare that certain entries are 'pending dequeue' as you propose.
- How do I do an efficient dequeue operation that is transaction protected
- and scales up to dozens of pending dequeues?
-
- As far as I can see, there is no PRACTICAL way to do this in straight SQL.
- Someone mentioned that PROGRESS' language can search for unlocked records.
- And no one yet has addressed the issue of how to WAIT on an empty queue,
- other than periodically polling, which (technically speaking) sucks.
-
- >
- >Well, considering that SQL is pretty much standard between Oracle and
- >all of our significant competitors I'm not going to run arround
- >making a fuss about how SQL can do the job. It's an industry
- >standard and we don't need to defend it any more.
-
- Weak, weak. "Industry Standard" translates to: "Gee, software is just too
- complicated to sell, so lets all agree on some simplistic solution and
- claim that doing things the right way is actually wrong, because its
- non-standard. We can sell a lot more software that way."
-
- >
- >As for "comp sci questions" versus real word questions, I really do
- >believe there is a difference. I'll admit that you've given a pretty
- >good example of a need for a FIFO queue, but in the vast majority of
- >applications you really don't need one. [...]
-
- Speak for your own real world applications. The ones I have been building
- for fifteen years REQUIRE queues. I suspect that if
- relational databases supported queues, they would be found to be widely
- useful. IMAO, relational databases will eventually be seen as a mass
- delusion, since they are not anywhere close to being a useful data model
- for the vast majority of applications. How many wordprocessors or spreadsheets
- or compilers or file systems use the relational model for their data?
-
- Jonathan Edwards
- IntraNet, Inc.
- I DO speak for my company!
-
-