- From: Martin Thomson <martin.thomson@gmail.com>
- Date: Sat, 29 Oct 2016 20:24:36 +1100
- To: Kari Hurtta <hurtta-ietf@elmme-mailer.org>
- Cc: Patrick McManus <pmcmanus@mozilla.com>, HTTP working group mailing list <ietf-http-wg@w3.org>
On 29 October 2016 at 16:49, Kari Hurtta <hurtta-ietf@elmme-mailer.org> wrote: > So on these case that heuristic that ignore immutable for > "weakly framed content" does not help either. 304 reply > still confirms to the client to keep using the corrupted content. > > My suggestion using immutable=<property from bydy> for ignnoring > immutable was making that more explicit than heuristic. But > that does change is that ignoring usefull or not. We aren't really looking for perfect here. Corruption happens, but is highly unlikely in the cases that we use this: Note that Firefox also requires HTTPS (I think). So random corruption is astronomically unlikely and well-framed responses (Content-Length fits, it uses h2) mean that there is enough confidence that the response was as the server intended. The key thing for the document is probably not to specify precisely what conditions that it would consider this safe to use, but to note that a degree of care might be needed to avoid having broken responses cached. Yes, the server might screw up, but the client won't be responsible for that. The server needs to create a new resource to fix it. Or maybe user will notice a problem and reload.
Received on Saturday, 29 October 2016 09:25:09 UTC