Re: [Acme] [Technical Errata Reported] RFC8555 (5771)

[Rob Stradling <[e-mail]>]

On 03/07/2019 13:14, Salz, Rich wrote:
>>     I don't think so.  IINM, HTTP caching only comes into play when a client
>      sends a subsequent HTTP request message that may (or may not) be
>      satisfied by a cached HTTP response message.
>    
> A client that follows caching shouldn't send a new request if the cache-control headers on the first response direct it not to, right?

Right, but the prerequisite is that the client explicitly wants to 
obtain a fresh (or at least unexpired) HTTP response message.  If that's 
not the case, then HTTP caching doesn't come into play, because HTTP 
itself doesn't come into play.

ISTM that once an ACME client has extracted a directory object from an 
HTTP response message, then both HTTP and HTTP caching are no longer in 
play.  The client is now only dealing with a directory object, not an 
HTTP response message.  (From Jacob's message, I get the impression that 
this is the prevailing understanding).

>      > Clients could also use If-Modified-Since, right?
>      
>      Good point.
> 
> Then maybe the errata could just be
> 	Clients SHOULD use an If-Modified-Since header to get more effective caching.

I think that's a useful optimization suggestion, but I don't think it 
addresses the issue.

The idea behind the erratum is to force HTTP caching rules to apply to 
directory objects, so that servers can update their directory objects 
and expect clients to take note.

-- 
Rob Stradling
Senior Research & Development Scientist
Sectigo Limited