[development] One-to-one tables considered harmful

[Earnie Boyd]

Quoting Chris Johnson <cxjohnson at gmail.com>:

> On 6/4/07, Earnie Boyd <earnie at users.sourceforge.net> wrote:
>
>> I have been reading the posts and wondering what is so different about
>> a comment.  A comment is nothing more than content very much similar to
>> a node.
>
> Well, not quite.  There is one very distinct difference that makes a
> comment very different from nodes as we treat them today.  And it's
> also the one thing that makes it hard to clearly call a comment a
> first-class object, in a modeling sense.
>
> That difference is:
>
> * nodes are "new" objects.  They "start" or "instantiate" a piece of
> content.  They are independent, typically.
>
> * comments are "follow on" objects.  They are a "reaction" to a piece
> of content.  They are *always* dependent upon the node to which they
> refer.
>
> So much for the model.  In a practical implementation sense, the
> questions are different, but need to keep the model in mind.
>

How is what you describe different from a child node dependent on a 
parent node?  If I have a content type relationship mapping table that 
maps a parent to a child then the children become the ``follow on'' 
objects.

Further more we can design the content type relationship mapping table 
such that it maps any content type to another content type.  So I can 
have one content type represent albums and the content be generally 
about the album and another content type represent songs whose content 
is about the song and a child of the album.

Yes, this is different thinking and I am asking that we think of a 
comment as nothing more than more content.  I am asking that all 
content be thought of as having possible relationships to each other.  
IMO, it will make for a more robust API and potentially less code.  
Obviously others think it is more robust because they build replacement 
modules for the existing core comment module.

Earnie