- From: Mark Nottingham <mnot@mnot.net>
- Date: Fri, 27 Aug 2021 12:22:12 +1000
- To: Martin Thomson <mt@lowentropy.net>
- Cc: ietf-http-wg@w3.org
> On 27 Aug 2021, at 12:19 pm, Martin Thomson <mt@lowentropy.net> wrote: > > On Fri, Aug 27, 2021, at 12:16, Mark Nottingham wrote: >> This doesn't change the behaviour so much as it adds to it, I think. > > An addition is also a change. Anyone that comes to depend on the trailer-based behaviour would be more dependent on the CDN that supported that behaviour if alternative providers did not. That's the source of my concern. Of course, CDN-proprietary extensions likely fill this gap, so I recognize that not doing this is not necessarily a real solution. As I said, we wouldn't define this for CDN-Cache-Control; it's only for fields that explicitly opt into it. For example, if BunnyCDN wanted to support trailer updates, they could say so in their documentation for (presumably) BunnyCDN-Cache-Control. The question here is whether *they* define it in that document completely -- with the possibility of mismatches with other implementers who make that choice -- or whether we define it here so that they can reference the standardised (yet optional, opt-in only) behaviour. -- Mark Nottingham https://www.mnot.net/
Received on Friday, 27 August 2021 02:22:33 UTC