Tool for optimizing extensions/packages (idea)

[sm8ps]

The following is inspired by the current thread slow boot, xorg-intel questions... but I want to keep it separate so as not to mix two different things.

I am open to letting us look into what can be left out by default for the average user.  The libLLV* one is large and I can't find anything in my huge sce that needs it according to ldd.

I have been thinking in a different direction: how about a tool that determines an optimal sub-set of packages to be included into a base extension? That could be a complementary strategy to trimming seemingly unnecessary files.

Some packages are  included into nearly each and every extension. The tool would have to determine a "weighted intersection" taking into account frequency of occurrence and file size of the package. Then it could make a suggestion on what to include into a base extension for different savings in disk space.
Like a package of 15 mB being included by 80% of all extensions would save 12 mB per package "on average" whereas one of 20 mB being included into only 10% of all extensions would only save 2 mB. Thus the latter would be less likely to be included into a base extension unless the user decided that some of the extensions making use of it were so frequently used that it would make sense nevertheless.
Finally, the tool could create the extension list and a script for sce-import statements to re-base one's extensions.

The tool needs to know which extensions are available (from 'tce/sce/') plus their dependencies (from 'DEBINX'?) and size thereof. I presume the basic tools are available from the logic of ''sce-import''. The complexity could grow by quite a lot, nevertheless I believe this would be a valuable add-on to optimize a dCore installation.

Opinions or comments welcome!

[nitram]

Like the idea. Maybe not too hard for a coder, script mounts all extensions and reads all /usr/local/sce/sce_name/*.md5sum files to determine frequency of each package across SCEs, might be able to calculate weighted average from recycled  sce-import -s  code, outputs results or creates custom list file. The only drawback i see is that it's based on a fully installed system, so all extensions would need to be re-imported dependent on the list file SCE.

[kevinfish]

I'm a total noob to dCore but what I don't understand is why sce-import can't just use debian's own package dependencies to create all the dependent sce packages at the same time it creates the called for package (and adding all those names to sceboot.lst if the -b flag is given).   Maybe the user should have the option to do it the way its currently done or do it this way depending on whether he wants a more straightforward configuration or is short on disk/thumbdrive space?