haskell-platform VS data-category

The original link http://www.haskell.org/platform/ does redirect to the downloads page, but http://www.haskell.org/platform (without the slash) doesn't. Still, I think it's probably better to make it explicit to visitors that the Platform is deprecated in favour of the Downloads page: https://github.com/haskell-infra/www.haskell.org/pull/167

The link for the official page (http://hackage.haskell.org/platform/) gives you a 404 error and every other link (including those on the official Haskell Platform Github https://github.com/haskell/haskell-platform) leads you to the generic downloads page on the official website which has absolutely no hint about Haskell Platform.

If you plan to try some of the examples, I would highly recommend setting up Haskell locally on your machine. You can follow this guide to download and set up ghc and ghci, we will use ghci for all examples.

Why does the haskell platform (https://www.haskell.org/platform/ ) require it in order to install the IDE?

As for GHC, check out this page for step-by-step explanations of installation. Personally I think it's much easier to set this up on linux compared to windows. I use Ubuntu WSL2 on Windows, this avoids a lot of the trouble.

Is it about the time to redirect the "Download Haskell" link in the right sidebar here from https://www.haskell.org/platform/ to https://www.haskell.org/downloads/?

If you save the sliced model as x3g (default for FFCP) and you get an empty file, follow the steps described in this issue to install the required libraries: https://github.com/haskell/haskell-platform/issues/231#issuecomment-177530716

Even a fairly simple statement like "F preserves direct limits over N" is basically impossible to express like this. You can step further from Hask and work at the type level until the very end (which I believe is the approach taken by data-category), you can resign yourself to only expressing things that can be "defunctionalized" (an appropriate use of the term, I think, if not a correct one) down to Haskell functions, which gets you (Co)Yoneda, Lan, Ran, etc. in the general case and I think Traversable in this particular instance, or you can take some intermediate approach with constrained functions and/or explicit witnesses in your data types, but you can't make proper category theory "just work" the way it should.