The Interesting Installation Integer Problem

[Paulo Casanova]

Max Kanat-Alexander wrote:

>	Now, what if we had a database system where all variable-length text
>fields had the same underlying implementation? That is, where varchar ==
>tinytext == mediumtext, etc. We couldn't trigger a change on moving
>something from varchar to text.
>  
>
And why would that be a problem? If all datatypes are equal, no change 
is needed in the underlying database... I don't get it...

>	(2) Store the entire current schema itself in the database. That way
>we'll always know what the "generic" type of a field is SUPPOSED to be,
>even if it's not possible to read out that information.
>	Problem: That's a somewhat-decent amount of work.
>  
>
Sounds fine. However, PLEASE, keep it manageable for people with custom 
hacks. Say, the checksetup script could complain about unknown tables 
and unknown fields but would not change them. As a rule of thum, I'd say 
checksetup should not drop any fields or tables unless we really know 
what we're doing (checksetup specific code?)

You know, dropping stuff in a database is kinda dangerous :)

Paulo