home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.databases
- Path: sparky!uunet!cs.utexas.edu!sdd.hp.com!wupost!darwin.sura.net!mlb.semi.harris.com!controls.ccd.harris.com!yaseen
- From: yaseen@controls.ccd.harris.com (Rahim M. Yaseen)
- Subject: Re: Object What is an OODBMS?
- Message-ID: <1992Aug14.124148.15752@ccd.harris.com>
- Originator: yaseen@rs2
- Keywords: OODBMS, C++, POET
- Sender: yaseen@controls (Rahim M. Yaseen)
- Organization: Harris Controls
- References: <RCZ5NS@netmbx.netmbx.de> <1992Aug13.170024.11525@genghis.borland.com>
- Date: Fri, 14 Aug 1992 12:41:48 GMT
- Lines: 42
-
-
- In article <1992Aug13.170024.11525@genghis.borland.com>, cy@genghis.borland.com (Cy Shuster) writes:
- |> [discusses limits of the relational model to store real world
- |> relationships, such as many-to-many]
- |>
- |> Have we really come full circle, back to Codasyl networked structures?
-
- In a sense, yes and no.
-
- The yes applies to the fact that networked structures would appear to be
- similar to the physical (storage-level) structures for OODBMSs.
-
- The no applies to the fact that in Codasyl databases, the user-level
- abstractions of relationships is very close to that at the physical-level
- (i.e., parent-child) thus requiring users to think like programmers.
- In OODBMSs, there is usually a user-level abstraction of relationships
- and corresponding semantics of such relationships (e.g. IS-A, IS-PART-OF,
- superclass-subclass, inheritance, etc. etc.). Several semantic data models
- capture such relationships e.g., extended ER, NIAM, Express, ..
- The argument is that if one knows the static relationships ahead of time,
- go ahead and define them (note that nothing prevents these relationships
- from being re-defined or changed later i.e., schema evolution)
-
-
- |> What about the vaunted features of ad hoc queries and on-the-fly
- |> joins, over casting relations into concrete at schema design time?
- |>
-
- The way I see ad-hoc queries and on-the-fly joins is that dynamic
- relationships are being created at run-time using domain-compatible values.
- There is nothing preventing a query language in OODBMSs from querying
- pre-defined relationships AND creating dynamic on-the-fly relationships.
- (I have seen a couple of experimental query languages attempt to do this)
- I see this concept as derived relationships: a concept similar to that of
- derived data.
-
- Thus, you have the advantages of two worlds: static (pre-defined) relationships
- (which are definable in many large applications) and dynamic (on-the-fly)
- relationships.
-
- - yaseen
-
-