Debian Pure Blends

Derivatives done right

LinuxTag, Berlin, 11. June 2010

Abstract

Large Linux distributions like Debian or Ububtu are typically featuring a flat pool of software packages which contains more than 10000 packages with the tendency to grow even larger. This constant increase of size without any visible substructure leads to helpless users who might have problems to pick the relevant pieces for their work on one hand and on the other hand it is hard to assure the quality of packages which might belong to a certain work field.

The Debian Pure Blends effort identifies packages which are useful for certain real life tasks, makes them visible inside the flat pool of packages using certain techniques and provides tools to make the QA work of the dedicated Blend team easier. In that way a substructure inside a large distribution can be established which enables users and developers to cope with the constant growth. Quite successful examples of Blends are Debian Edu, Debian Med and Debian Science.

The Pure Blends approach can also be seen as "derivative done the right way". What does this mean: Out of all large distributions a lot of smaller distributions have grown. Some of them are successful but most of them end up completely unmaintained and outdated after some time when derivers might have lost interest. Also the derivative is out of the strong QA measures of the mother distribution and needs to invent a completely separate infrastructure. The Blends approach however, reduces the amount of work to people who would like to adapt a Linux distributions to a minimum. They just have to find out what is missing to solve a certain task and adapt the mother distribution (the source) to the purpose they want to fullfill. The basic idea is: Do not make a separate distribution but enable the distribution itself to fit your purpose best.