createdbyvictor / whirling-canary Goto Github PK

Have a more physics based rotation instead of the existing linear implementation.

Some sort of reliable performance benchmark detection for perceiving client-side performance.

All we have right now is the ability to detect whether or not features are available to us in a browser and also kinda sorta what browser client we are on as well as its version. Because this solution requires rendering on the client, we need to be able to detect whether doing so will provide a poorer performance than delivering regular-flat images or even a simple single image. All of this depends on the device that he user is using, which we currently don't know anything about.

WebGL is currently not in use because the image composition techniques found in Canvas2D appear to out-perform it for the same task. Further Research should still be done. Also keep track of WebGL2, which at the time of writing has just landed on Firefox and Chrome but currently not supported on mobile browsers.

Offloading the graphics manipulation tasks directly to the GPU is the ideal here.

This may be a far future thing. Unfortunately each browser has a limited amount it will allow for storage per domain. There is no overriding it that I have found so-far. Also, a single product's rendered images could be about half a gig. Which is nuts.

We will need to incorporate this into the zoom functionality as well.

Make it quick easy for others and ourselves to pick up the project and run with it going forward. Should aim at removing as much folklore as possible.

How should we deal with users that use an older IOS version whose browser disables javascript on scroll?

Listing what the user can do and how they can do it.

Needs to be done before we have customers.

Most of the work is already done for this, we just need to put in the effort to make it cross-browser compatible.

I don't know the details, just that it is something to think about for the future.

Test whether an image which is the exact size of the image tag, will trigger an ImageDecode event when you quickly swap in and out new values to its source tag. Not triggering the ImageDecode event is the goal, as it makes the FPS drop. Using canvas allows us to avoid it as it dumps the pure image-data into memory.

Currently they show nothing. at all. it's dumb.

This is a Day 2 feature. The ground work as already been laid but we need to decide on an interface for the users to interact with. Pinch to zoom makes sense, but how about on browsers which don't have a touch screen?

Also, work needs to be put in to make non-canvas solutions work with however we decide to make this work.