Oh man, that’s like orders of magnitude more than what was needed to fly to the moon!
Is it before xz compression or after?
It is weird that you need all this code to “just” validate the received JSON. On another hand it’s not THAT much compared to a lot of other things we’ve been adding to Blender.
If it allows us to have a better quality and easy to develop online aspects of Blender, it probably worth it.
But still asking myself whyyy its so big 
Sure, otherwise we’d not consider it, right 
What I meant is that maybe with some reduced amount of requirements to the parametric search of libraries we can get some other alternatives we can look into.
Like, if we say: validation is essential, http codegen not so much, then maybe there is a library that we can just add without too much discussions about it because it’ll be so small.
The point about generating boiler plate for converting JSON to real objects is valid, but maybe we can do it in less than 11 megabytes of code, because, how hard could it be to just assign some properties and arrays once the schema is validated?
If dependencies are needed at runtime then indeed is not SO relevant where the YAML file is stored.
Pydantic is not that big though, the wheel is 456 KiB, unpacked is about 1.6 MiB.
What are those 11 MiB consists of?