![]() | |
#11
| |||
| |||
|
|
I agree with both Jawn and dataman. But, I can tell you that sometimes, what seems to be something that can easily be designed directly as a relational schema (skipping the ERD), can come back to bite you. The ERD helps you see relationships in the data that you may miss otherwise. In our case, we missed that there was a specialization/generalization. We had to rebuild. It is cheaper to do it right the first time than to go back and rebuild after you found that something doesn't work the way you need it to. Coming at it from a non-relational bias ... :-) |
|
"John" <?> wrote in message news:3fcee2f7$0$14053$fa0fcedb (AT) lovejoy (DOT) zen.co.uk... I was taught about ER modelling at University but found it to be a pointless abstraction, preferring to just write down the schema directly. What does the group think of ER modelling as a design approach, and what alternatives do you use? John Cheers, |
#12
| |||
| |||
|
|
"dataman" <dataman (AT) ev1 (DOT) net> wrote in message news:vsv181b4k0269 (AT) corp (DOT) supernews.com... "John" <?> wrote: I was taught about ER modelling at University but found it to be a pointless abstraction, preferring to just write down the schema directly. What does the group think of ER modelling as a design approach, and what alternatives do you use? John I don't view it as a waste of time. I've not found a better way to identify dependencies in data than logical data modeling. You may have confused logical data modeling with database design. " They're really two different things. " Maybe that's the issue. The data modelling bit seems plausible to me, and I am happy enough drawing the diagram. I am also happy enough designing a database, it's just that when I am doing E/R modelling, I find myself trying to second guess the abstraction in order to ensure that the eventual relations are sensibly defined. When I'm using it in that fashion, it does seem pointless. Date (IDS), for example, sees the E/R model as a "set of informal concepts", while "relations are formal objects, and the relational model is a formal system". He also (sensibly) suggests that when Codd constructed the relational model, the "useful semantic concepts" he was using were "basically ... those of the E/R model". This view seems to support my belief that E/R modelling is fundamentally informal, something which questions its very status as a *model*. Perhaps I just have problems with informal / wooly / subjective things. I suspect that's why I'm a computer scientist rather than an artist! John Lots of folks do find documentation pointless. In fact some found it so |
#13
| |||
| |||
|
|
"John" <?> wrote I was taught about ER modelling at University but found it to be a pointless abstraction, preferring to just write down the schema directly. Me too. An ER drawing is not a complete database design like a ground plan is not a complete building design. A database design include many aspects (integrity constraints, derivation rules) that an ER drawing can not express. What does the group think of ER modelling as a design approach, and what alternatives do you use? I create the logical database design using written Tutorial D. You can accompain the logical database design with one or several ER pictures for clarification purposes, but the ER pictures are only pictures. A picture is only one part of a physical data model using a tool like ERwin. |
#14
| |||
| |||
|
|
You can accompain the logical database design with one or several ER pictures for clarification purposes, but the ER pictures are only pictures. A picture is only one part of a physical data model using a tool like ERwin. You also have all of the associated meta data for that data model (assuming you choose to enter it). That includes integrity constraints and derivation rules. |
#15
| |||
| |||
|
|
"dataman" <dataman (AT) ev1 (DOT) net> wrote You can accompain the logical database design with one or several ER pictures for clarification purposes, but the ER pictures are only pictures. A picture is only one part of a physical data model using a tool like ERwin. You also have all of the associated meta data for that data model (assuming you choose to enter it). That includes integrity constraints and derivation rules. Using which language? Is the language relationally complete? Regards Alfredo |
![]() |
| Thread Tools | |
| Display Modes | |
| |