Re: Does Codd's view of a relational database differ from that ofDate&Darwin? [M.Gittens]

paul c schrieb:
> Alexandr Savinov wrote:
>

>>  
>> ... If something disappears, if something is deleted then acutally we 
>> get null.

Actually it is one and the same, i.e., null, false, empty set (may be you can continue) are different terms from different sciences for one and the same. Of course, there might be variations in interpretations but generally the purpose and the role in the theory are the same.

> i'd like to know what problem the never-ending arguments about nulls are
> aimed at. in mundane applications at least (which i think is where most
> people spend their time), what is the problem with using empty strings
> for unknown names or zeroes for unknown number values? if the db user
> has allowed tuples with such attribute values then they should expect to
> have to judge the results accordingly, such as ones produced by
> aggregate functions. if they want to record some other kind of
> 'meta-data' such as keying problems, then they can define relations
> specific to those.

1. DBMS has a very concrete interpretation of NULLs which is encoded in the database logic of bechavior, say, aggregation funtions or joins.
2. users are allowed to use NULL values for any other purpose they like.

The problem appear when the interpretations of the system and the user do not conicide. This means that the user expects one result while the system returns another result.

One solution: do not use NULLs at all.
Another solution. Use it only as the system supposes it has to be used.

-- 
alex
http://conceptoriented.com