[css-fonts] Investigate using will-change: content to decide whether to use incremental font downloading

[emilio]

As per the discussion in #6063.

cc @litherum

[tabatkins]

Oh, I'd assumed we were discussing another will-change value that was merely like content in that it didn't refer to a property.

[svgeesus]

@emilio is this issue still relevant (given that IFT has changed a lot since it was opened).

If so, could you explain a bit more how will-change affects the font download decision?

[emilio]

I think so? Is there a way for the page to annotate that the content is likely to change and maybe you want to do less fine-grained incremental downloads?

[svgeesus]

maybe you want to do less fine-grained incremental downloads?

Ah. With the old patch-subset approach, the client provided a list of

• what it could render
• what it wanted to render

and the server computed a patch for exactly that. But this was a cache-buster, required a special server, and was a potential privacy leak.

With the current approach, the initial font contains tables of (urls to) patches, along with what is in each one. The client doesn't communicate with a server except to request some patch files. Of course the client can request patches for more than it currently needs; and as the patches are pre-computed (and thus, the same for everyone, and cacheable) they tend to not be very fine grained.

But still, it could be a useful hint for the client.