Re: VCDIFF

VCDIFF does have the advantages of being well-documented and quite straightforward. A Brotli-only could mean, de facto, a single implementation given how complicated it is. That's something worth keeping in mind.

Skef
________________________________
From: Garret Rieger <grieger@google.com>
Sent: Tuesday, January 17, 2023 12:22 PM
To: w3c-webfonts-wg (public-webfonts-wg@w3.org) <public-webfonts-wg@w3.org>
Subject: VCDIFF


EXTERNAL: Use caution when clicking on links or opening attachments.


The patch subset method currently specifies two options for compression of the patch: VCDIFF and shared brotli. Shared brotli produces smaller patches than VCDIFF. Originally VCDIFF was included as an option since shared brotli was in it's early days so there wasn't a public implementation or finished specification. The idea being that if shared brotli wasn't ready yet implementations could be built with VCDIFF instead.

However, that's since changed and we have a open source implementation of shared brotli now, and we have a workaround for the specification being a draft (see: https://github.com/w3c/IFT/issues/101#issuecomment-1251591398<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fw3c%2FIFT%2Fissues%2F101%23issuecomment-1251591398&data=05%7C01%7Csiterum%40adobe.com%7C754408f293ff4b5a708808daf8c8a8df%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C638095837957266676%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=I1K%2FID4p4sffO9HUckz7w1Qk1hVigYn7Cr5cEcPgkUQ%3D&reserved=0>). Also, we've demonstrated that shared brotli is sufficiently performant for use (see: https://docs.google.com/document/d/1MdtB_WPC2grAx3vFgLHA1-CqRzQtWj_cJYx40W2VQgI/edit#heading=h.dy3wd9u6su7<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdocs.google.com%2Fdocument%2Fd%2F1MdtB_WPC2grAx3vFgLHA1-CqRzQtWj_cJYx40W2VQgI%2Fedit%23heading%3Dh.dy3wd9u6su7&data=05%7C01%7Csiterum%40adobe.com%7C754408f293ff4b5a708808daf8c8a8df%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C638095837957422901%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=JUNyUhRzD%2BfI1SuH%2BAZkt%2Fu0nPwUYdG05iVpsHkMCRg%3D&reserved=0>).

Given all that I wonder if there's still any value in keeping VCDIFF as an option in the specification? The advantage of removing it would be reduced complexity of implementing the specification (both client and server) since currently VCDIFF support is mandatory. Let's discuss this during the next working group call.