Add endpoint for image proxying

[mickvandijke]

Torrent descriptions can contain externally hosted images.

To preserve our users' privacy, we should load external images through a proxy.

[mickvandijke]

Hey @josecelano, @da2ce7

I was thinking of adding an image cache to the backend and a public API endpoint that takes a direct image URL as an urlencoded GET parameter. The API endpoint to be called /imageproxy would then fetch the image from the ?url GET parameter on request, then save it in a cache and serve it back to the client.

Example:

We have an <img> on the frontend with src="https://en.wikipedia.org/wiki/Test_card#/media/File:SMPTE_Color_Bars.svg". This is an external source, so we are going to load the image using the backend's /imageproxy endpoint.

To pass the image src as a GET parameter to the /imageproxy endpoint, we would first have to url encode it: https%3A%2F%2Fen.wikipedia.org%2Fwiki%2FTest_card%23%2Fmedia%2FFile%3ASMPTE_Color_Bars.svg.

We then substitute the img src with the imageproxy link: src="https://api.torrust.com/imageproxy?url=https%3A%2F%2Fen.wikipedia.org%2Fwiki%2FTest_card%23%2Fmedia%2FFile%3ASMPTE_Color_Bars.svg".

This should then display the original image, but through our proxy.

What do you think of this approach?

[josecelano]

Hi @WarmBeer, as we discussed, I see these prons and cons:

Prons

• We do not need to handle media uploads.
• We do not need to keep the images. The cache will be filled again as users use the application. That makes the deployment simpler, although we need a storage anyway for other things like the SQLite database.

Cons

• It could be slow the first time we serve the image because we have to get it from the original location.
• It could be inefficient because we need to get the image again when the cache expires. We need to check again that the image is an actual image.
• The original image link could be broken (something we have to warn users about).

Notes

• Check that the image is an actual image (not only with the content-type: image/png header).
• Use a two-level cache: memory and disk. For the time being, we can only use disk cache.
• Allow the browser to cache the image (with the right response header). WE can use versioning to update the image if the original image changes, but I do not think we need that now.

[da2ce7]

IndexMap could be very good for a in-memory cache. You can set the max-items of the cache, and insert/pop items once it reaches the desired file. (This works if the cache has some limitations of the size of the items).