URI for generated image resources

[IS4Code]

Has anyone considered or thought of making a URI scheme specifically for Stable Diffusion?

For example if I have this image with these properties:

Prompt: train, snow mountain, sunset, (masterpiece),((ultra-detailed)), (highly detailed CG illustration),(expressionless), (best quality:1.2), High quality texture, intricate details, detailed texture, High quality shadow, Cinematic Light, Depth of field, light source contrast, perspective,20s, (ulzzang-6500-v1.1:0.5), <lora:epi_noiseoffset2:1>, <lora:add_detail:0.8>,
Negative prompt: EasyNegative,paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2), lowres, normal quality, ((monochrome)), ((grayscale)), ng_deepnegative_v1_75t
Sampler: DPM++ 2M SDE Karras
Model: majicmixRealistic_v7
CFG scale: 9.5
Steps: 40
Seed: 500179606
Clip skip: 2

I could imagine a URI of this form (shortened and added newlines):

sd.gen:
?prompt=train%2C%20snow%20mountain…%3Clora%3Aadd_detail%3A0.8%3E%2C
&neg_prompt=EasyNegative%2Cpaintings%2C…ng_deepnegative_v1_75t
&sampler=DPM%2B%2B%202M%20SDE%20Karras
&model=urn:sha256:7C819B6D13663ED720C2254F4FE18373107DFEF2448D337913C8FC545640881E
&lora.epi_noiseoffset2=urn:sha256:…
&lora.add_detail=urn:sha256:…
&cfg_scale=9.5&steps=40&seed=500179606&clip_skip=2
&width=1152&height=1664

The purpose of this URI is to store together all parameters that uniquely identify the image, i.e. the dimensions, seed, the hashes of the model but also the individual LoRAs. There may be other parameters that affect the generation (VAE?), and as far as I understand it, the particular prompt format (i.e. the meaning of parentheses, <lora etc.) should also be identified in some way.

Stable Diffusion web UI or any other compatible software could then register the sd.gen URI scheme, and when a browser navigates to any such link (which sites like Civitai would offer), the software could easily load all the components, or warn if any required files are missing. Some basic prompt rewriting would be nice too, for example this image uses LoRA epi_noiseoffset2, but if the same model had a different filename in a particular installation, it could automatically notice <lora:epi_noiseoffset2:1> in the prompt and rewrite it to match the local filename.

This could work not only for txt2img but also img2img, where the source image (and mask) could be stored as another URI parameter, as a hash, local file name (file:///) or http URL.