Path: text.usenetserver.com!out03b.usenetserver.com!news.usenetserver.com!in04.usenetserver.com!news.usenetserver.com!nx01.iad01.newshosting.com!newshosting.com!198.186.194.249.MISMATCH!news-out.readnews.com!transit3.readnews.com!207.99.111.53.MISMATCH!newspeer1.nac.net!news.astraweb.com!border2.newsrouter.astraweb.com!not-for-mail From: Tim X Newsgroups: comp.databases.theory Subject: Re: Modeling question... References: <4873acc6$0$4052$9a566e8b@news.aliant.net> <4874aba0$0$4069$9a566e8b@news.aliant.net> Date: Sat, 26 Jul 2008 16:09:56 +1000 Message-ID: <87r69hp45n.fsf@lion.rapttech.com.au> Organization: Rapt Technologies User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.0.60 (gnu/linux) Cancel-Lock: sha1:+WoEKJRxarrVcF8yPzKHkPaplwA= MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Lines: 63 NNTP-Posting-Host: d76df1ff.news.astraweb.com X-Trace: DXC=1i8P<3I3V2:hZ4>`7BGPo4L?0kYOcDh@:IGb]mONgOi:bTVG84H7a:0RTFYYW29Y\:eIa1VJ6=Ra7 Xref: usenetserver.com comp.databases.theory:172134 X-Received-Date: Sat, 26 Jul 2008 02:10:02 EDT (text.usenetserver.com) "David Cressey" writes: > "Volker Hetzer" wrote in message > news:g6cmjq$9s9$1@nntp.fujitsu-siemens.com... >> Bob Badour schrieb: >> >>> Ooooh! Reinventing EAV with levels... >> >> >> >> Possibly. I had a look at >> >> http://ycmi.med.yale.edu/nadkarni/eav_CR_contents.htm and didn't find >> >> anything exciting. >> >> All my attributes (key value pairs) are (for the purpose of this >> >> discussion) strings, so the Data tables hierarchy ends with >> >> EAV_Objects in the first image of that link. >> >> My problem is that, that I haveTest three different "Objects_1" >> >> tables and I'd like to avoid having to replicate the EAV_Objects-Table >> >> for each "Objects_1"-Table. >> >> OTOH, I could have the "level" entities all be children of an id table >> >> and >> >> put the key value pairs into a child of that table. I need to try this >> >> out. >> >> >> >> Thanks for providing the pointer! >> >> Volker >> > >> > Just to be clear, I was more than offering a pointer. I was also >> > ridiculing the idea of EAV. >> I got that. :-) >> But "we want to be able to create and delete attributes" is a customer >> requirement. I think it's different from "I am too lazy to do a proper > data >> model". There are plenty of "normal" attributes left to model ERD like. >> > > Here's the problem with the customer requirement that "we want to be able to > create and delete attributes": the same customers nearly always want to be > able to derive value from the data in the database by using it with standard > queries to drive standard reports, charts, or extracts that can be fed to > another process. After all, that's inherent in database data, isn't it? > Well, not if the customers can each invent their own attributes, it isn't! > > If you've never been down this road before I suggest you stay in the same > job long enough to see what happens when you tell the customers that each of > them has to design their own queries, because each of them defined their own > attributes. What happens isn't pretty! I would have to agree very strongly here with David's warning. I also suspect that the customer's 'requirements' may actually represent a misunderstanding on their part or at least poorly expressed requirements. I would strongly recommend digging deeper and finding out more about why they feel this requirement is necessary. It reminds me of many clients I have had who bring me a solution to implement rather than bringing me the requirements and letting me come up with a solution in conjunction with them. This is a frequent and difficult problem faced by developers. You need to dig deeper and understand why they believe they need the ability to add/drop attributes. regards, Tim -- tcross (at) rapttech dot com dot au