[6601] in www-talk@info.cern.ch
Re: An alarming trend in METHOD=POST
daemon@ATHENA.MIT.EDU (James Pitkow)
Wed Nov 9 21:12:33 1994
Date: Thu, 10 Nov 1994 02:25:37 +0100
Errors-To: listmaster@www0.cern.ch
Reply-To: pitkow@cc.gatech.edu
From: pitkow@cc.gatech.edu (James Pitkow)
To: Multiple recipients of list <www-talk@www0.cern.ch>
Hello,
Chris Lilley, Computer Graphics Unit on Wed Nov 9 05:42:24 1994 wrote:
> > Can POST really be cached or added to a hotlist correctly?
>
> No. That is the point. Things which should not be in a hotlist use a method
> other than GET, so they can't be put in a hot list. That is, I believe, what
> TimBL was saying.
One of the points made earlier was that caching/hotlisting heavily
paramterized GET commands can be undesireable - i.e. they tend to bump the
file sizes of hotlists, global histories, etc. I'm glad that you also
agree that hotlisting POSTs is worrisome - though, they are doable - just
store the data blob along with the URL in a structured format. Still, the
loci on control determining allowance of this ought to ultimately rest with
the user.
Conclusions from discussions:
use GET for logical independant tasks/retrieval
currently POSTs should not be hotlisted/cached, though this is doable
session-based transactions suit interactive tasks whereas stateless
transactions suit most paramterized IR tasks
Any others?
Jim.
