[Mentalunix-developers] mpkg package database stuff

[Clinton E. <men...@cr...>]

I had a bit of an idea at school today about mpkg's package database -
- why go through all the work of setting up our own package database, 
when we could use an abstract interface to any number of db's. Let me 
explain.

Ok, step one would be to write a library with database access methods 
for mpkg. Example: find_file(), find_pkg(), etc, that would be used 
when installing or finding packages to install(so the package manager 
would be able to find existing packages, available packages, and see 
what packages provide the files the package you are installing's 
dependencies,etc). Now, why write an entire database when all we need 
are these functions to access the database?

There are many free SQL databases(postgre,mySQL,etc) that can be 
searched and have fields added too, so why not use them? My idea is 
to put all the functions headers in a header file, and then make the 
implementation a nice shared object -- just pop in whatever interface 
you wanted. The first of these would be a SQL interface(postgre or 
mySQL?). The functions would just wrap around the sql server 
interfaces, providing a standard method to access the package 
database. The eliminates are a large amount of work on out part, and 
also adds flexibility. Don't want to have your pkg database in a SQL 
server? Fine, use a flat text interface. Don't like that? Then use 
any number of interfaces to different package database access 
methods. See how this is a good thing?

Implementing our own db format for the package database would be 
pointless and just add more work to the task of writing mpkg. So, 
does my abstract interface seem to make sense?

-------------------------------
#indrema @ irc.openprojects.net
be a part of the revolution.
   unknown_lamer