Right now we have universal height/width support, with possibility of specifying only one of these and original aspect ratio will be used. However, on images with very wide/tall aspect ratios this can be far from what UIs expect.
One strategy used is to say, give me an image where height is up to N and width is up to N.
Also, the semantics of height/width is to provide exactly the dimensions specified, making it not possible for us to return cached (already processed) very-similar-but-not-100% over.
A challenge is that currently we do assume a 1-to-1 mapping of request parameters and an output image. With loose constraints like this, there is a large number of outputs possible.
This can potentially help in consolidating image sizes requested by clients, which reduces costs, and improves responsiveness for users due to higher cache hit ratios.
Right now we have universal
height
/width
support, with possibility of specifying only one of these and original aspect ratio will be used. However, on images with very wide/tall aspect ratios this can be far from what UIs expect. One strategy used is to say, give me an image where height is up to N and width is up to N.Also, the semantics of height/width is to provide exactly the dimensions specified, making it not possible for us to return cached (already processed) very-similar-but-not-100% over. A challenge is that currently we do assume a 1-to-1 mapping of request parameters and an output image. With loose constraints like this, there is a large number of outputs possible.
This can potentially help in consolidating image sizes requested by clients, which reduces costs, and improves responsiveness for users due to higher cache hit ratios.