Modeling independent attributes on m:n links

Axel,

> That is certainly not an EAV design <
Indeed this was not a good comparison. My apologies if this introduced
some confusion.

To come back to Roberts question as to which design I would prefer
when (without boolean flags!). I can basically see two situations:

1) Some simple m:n relationships which just differ by some sort of
"relation_type" column and carry no further attributes with them. This
will fit well into one table (see Joes example)

2) Several independent relationships between the same tables which are
very different creatures (see the example I was giving) having
different business rules and their own attributes. They might not fit
well into one table, unless you get rid of most constraints and might
also be unable to assign a proper primary key. Further you would
introduce columns which apply to one relationship, but not to the
other. That was how I was understanding the initial post and therefore
was thinking at the mindset of "one-table-does-it-all" which also
drives EAV and OTLT/MUCT.

brgds

Philipp Post