Re: bad timing with r->request_time

["Roy T. Fielding" <fi...@liege.ICS.UCI.EDU>]

> I'm creating pages on the fly that are cached, but the cache files are
> often created a second or two after the request is received.
> 
> In "read_request", Apache records the time at the beginning of the request,
> then later in "set_last_modified" it uses that time to override the
> correct "Last-Modified" timestamp with:
> 
> ts = gm_timestr_822(r->pool, (mtime > r->request_time) ? r->request_time : m
> time);
> 
> This is a nuicance since I now have to refetch the entire cached document
> because the "Last-Modified" Apache gives me is wrong. The logic is
> there to prevent futuristic "Last-Modified" dates from being sent, but
> anything created as a result of the request is deemed to be in the
> future. Doh!.
> 
> I would suggest that "set_last_modified" be allowed to grab the time again
> so that this problem goes away while still preventing futuristic datestamps
> being sent to clients.

I think it more likely that r->request_time should be set closer to when
the response is generated.  I'll try to see if this is feasible.

......Roy