Re: What is this model technique called

"jimv" <member30950_at_dbforums.com> wrote in message news:2981867.1055186932_at_dbforums.com...
>
> I have been reading this thread with interest, since I was thinking that
> EAV would work well for my application. It does seem to be quite
> popular for storing medical records, as can be seen by a quick google
> search on EAV.
>
> I am working on an inventory system. However, I can assume very little
> about the type of data that is associated with inventory items. I know
> that every inventory item will have a manufacturer, model, and serial
> number, but I can't assume anything else.

Let's get back to jimv's example please; it's a better example than my original problem and (to me) seems to be a good fit for EAV. Another good example is BestBuy.com's site that allows you to compare product specifications for products of the same class. BestBuy gets either new product types or new specification attributes every week.

In both examples, there will be some very common attributes that should be modeled classically. However, there will be attributes (such as viewable size, refresh rate, watts per channel, gigabytes, printing speed, etc.) that will change very frequently as new products are created or the way they are described changes. These should still be captured so any that are "important" (i.e. for comparing products) can be reported consistently. Granted, enforcing this "attribute importance" cannot be done within the DBMS through constaints, but now becomes an audit function.

I understand that this type of design can be very dangerous and there is a real risk of it infecting the rest of the design. But then in some cases the alternative requires constant change to the data model.

Jarson Received on Fri Jun 13 2003 - 02:35:07 CEST