Today I’ve made a document, it’s resulting size was quite normal:

$ ls -s *odt 
5.9M Руководство по системе управления порталом.odt

All images are linked and reside under ‘img’ subfolder, which doesn’t take more than the document itself.

$ du -hs img/
2.1M	img/

And the resulting document takes… whooping 53 MiB(!).

$ ls -s Руководство\ по\ системе\ управления\ порталом.pdf 
53M Руководство по системе управления порталом.pdf

It was exported with PDF/A-a1 option, and I understand, that there are also fonts embedded into the doc, but that’s around 5 megabytes including all the variants (bold/italic), not more. Let’s export one image from a PDF viewer to disk and compare the size:

$ du -hb '/home/dtr/work/docs/website cms/platformiosedit.png'
1836974	/home/dtr/work/docs/alma tv/website cms/platformiosedit.png

$ du -hb '/home/dtr/work/docs/website cms/img/0250.png'
87942	/home/dtr/work/docs/website cms/img/0250.png

What?.. 

$ echo 'scale=1; 1836974/87942' | bc -q
20.8

Twenty, almost twenty one times bigger? What kind of archiving is that?
Let’s not use PDF/A-a1 and compare again

$ ls -s 'Руководство по системе управления порталом w\o PDF\A-a1.pdf'
2.0M Руководство по системе управления порталом w\o PDF\A-a1.pdf

Two megabytes! Two! As it should be. It is interesting to export that same iamge from the PDF viewer again.

$ du -hb 'platfromiosedit2.png'
91046	platfromiosedit2.png

Well, the size is bigger than that of the original image, but at least it’s close.

My question stays the same since #93462: why LibreOffice thinks, that it should modify original images, when it was clearly said to use Lossless images?