httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <>
Subject Re: [PATCH] ETags and Last-Modified, take 3
Date Tue, 02 Sep 1997 17:28:09 GMT
The script is killed after a timeout on client write too.  Scripts have to
expect to be killed.  They're broken if they don't expect it.  I don't
think there's a need to special case anything.


On Tue, 2 Sep 1997, Rodent of Unusual Size wrote:

> >From the fingers of Dean Gaudet flowed the following:
> >
> >On Fri, 29 Aug 1997, Alexei Kosut wrote:
> >>
> >> I don't think the script should be killed (some CGI scripts will do
> >> important things during/after sending an entity). Maybe send it a SIGURG or
> >> some other non-terminal signal, to tell it "hey, stop it, I'm not sending
> >> anything anymore"?
> >
> >It is killed off without reading anything if a HEAD request is performed. 
> >This is essentially the same ... isn't it?  I'll look at the patch later.
>     Grey-area time..
>     The HEAD method, and the appropriate action for processing, is
>     defined in both HTTP/1.0 and 1.1.  It seems reasonable for the
>     server to expect the script to have Done The Right Thing based upon
>     REQUEST_METHOD, and nuke it upon receipt of its headers.
>     The conditional stuff, though, is all HTTP/1.1 only (I think).  Is
>     it right for the server to expect a potentially 1.0 script to have
>     done the necessary calculations?  The script might think it's
>     supposed to continue, and be doing things which oughtn't to be
>     interrupted.
>     I think I'm leaning toward killing it and slurping anything it sent,
>     because scripts should be written with extreme-prejudice termination
>     in mind anyway.
>     How about this: we do conditional processing IFF the script returns
>     a "Status: 200" or no Status: at all?  If it returns anything else
>     we assume the script knew what it was doing, just pass it along, and
>     let things proceed normally.  If we did the conditional processing,
>     and it failed, we kill the script.  I think I see some flaws in
>     this, but just for discussion..
>     Maybe something that should go into the "new" CGI spec is a
>     script-response that says, "hey, I know HTTP/1.1".
>     #ken    :-)}

View raw message