- From: Daniel Stenberg <daniel@haxx.se>
- Date: Tue, 12 Mar 2019 10:53:32 +0100 (CET)
- To: Sebastiaan Deckers <sebdeckers83@gmail.com>
- cc: Jeffrey Yasskin <jyasskin@google.com>, HTTP Working Group <ietf-http-wg@w3.org>
On Tue, 12 Mar 2019, Sebastiaan Deckers wrote: > Is there a popularity cutoff % at which point we can break the web? As > stated, Refresh has 23+ years of real world use. Even ~1/5000 responses > (from a tiny 52M sample) is a huge amount on the scale of the web. There are so many dimensions to this! =) 1. I don't think *we* (as in the IETF or HTTPbis) would break the web at all since this header isn't something that is standardized or documented to begin with! I'm sure that no matter what we say here, the browser vendors (who seem to be the primary implementors of this) will have their own way of making this decision, and it will probably be enough with one or two fairly high profile sites to use the header for it to remain for a long time to come. Unless someone works with those sites. 2. I can't find a lot of support for this header outside of the browsers, so this header is already to some extent "breaking the web" by being used. 3. Since it isn't documented anywhere, it's hard to implement support for it in a way that doesn't introduce yet another version of dealing with it. So it's not clear what way that breaks the web more: leaving it unimplemented or implementing it in a subtly different way than how it already works elsewhere... BTW, my request count was done on requests done to ~52 million different origins out of which exactly 9080 returned a Refresh header. That of course doesn't exclude that the header might be used on other origins too if another request had been made. -- / daniel.haxx.se
Received on Tuesday, 12 March 2019 09:53:57 UTC