https://forum.world.st/Cuis-base-image-reduced-by-25-tp4695195p4696406.html
> non-essential functionality into .pck.st files. The result is that I
> could reduce Cuis by about 25%. Now it is below 500 classes and below
Congratulations. Amazing!
>
> - Compression
> - Graphic-Files-Additional (only png and tiff, as bmp and jpg are in the
> base image)
> - Sound
> - Theme-Themes (Only the default class Theme is included in the base image)
> - Network-Kernel
> - LinearAlgebra
> - Test
>
> If you update your image, all this code will be removed. If you download
> the updated image, it is no longer there. The packages themselves are
> also available at github, together with a txt file with a possible load
> order. I hope you'll find this more convenient when contributing code.
>
> There are several open questions, and I ask for your opinions and thoughts.
>
> 1) The "official" Cuis image can have none/some/all of these preloaded.
> We need to allow for clean unload of packages, but if there are no
> overrides, I think it is not a big deal. I'm not sure what's best. Maybe
> all these packages (and maybe others) should be preloaded by default.
>
> 2) I'm starting to believe that it is best to have those packages that
> are kept up to date in the same git repo as the base image. This will
> make it very easy to know what is the proper snapshot of some package to
> be used with a particular Cuis update level. Besides, it makes sense to
> have a Cuis-Smalltalk organization. Therefore I created
>
https://github.com/Cuis-Smalltalk , although I haven't saved anything
> there yet. I'd like to discuss with you the repositories to keep and
> their structure. I think that we might store there those packages that
> are preloaded, and those that are well maintained and actually used. But
> on the other hand, folks wanting to have better control of their code
> might prefer to host them in their own repositories...
>
> After all this,
https://github.com/jvuletich/Cuis should be deleted or
> marked as historic.
>
> BTW, there are a few other fixes and enhancements in this last batch of
> updates. #reduceCuis now produces a 3.6Mb image with less than 80kLOC.