Comments on SQL Backend

To all,

Comments on SQL backend for current.

When I first wrote it up, I used MySQL, the world's fastest mostly free
SQL server, even if it is "simple".  MySQL does not support foreign
keys, cascading deletes, sub selects, Stored Procedural Languages, or
Triggers.  But it is fast!!!  Hunter liked PostgreSQL instead, so I
converted it.

Sure, there is room for improvement in my partial implementation. But no
major changes are required in the database design. Most of the remaining
work is in the python modules interfacing to it.  As new tables are
needed, they will be added. If we need to enforce data integrity, we
can. If we need a new index, we'll add it.  If we don't need something,
we'll delete it.

What Todd has proposed is not much different than what I have already
implemented.  It uses some capabilities that MySQL did not have (as
listed above). It uses alternate syntax to accomplish much the same
thing (like implicit indexes). It does a mass table/field renaming.  It
has fields that will not be used.  Some fields are in the wrong tables.
It probably has indexes that will never be used. It makes the assumption
that the world runs on i386. But thats OK. My prototype is not perfect
either.


Two more "Rules"
11. No changes to the up2date client.  Since the client is written in
Python, one could build off it to automate things like automated
registration if one needs to.

12. Server can serve up more than one canonical architecture:  alpha,
ia64, sparc, i386, ....


Cheers,

Toby D. Reeves