- From: Roy T. Fielding <fielding@liege.ICS.UCI.EDU>
- Date: Thu, 29 Aug 1996 16:00:55 -0700
- To: jg@zorch.w3.org
- Cc: Larry Masinter <masinter@parc.xerox.com>, mogul@pa.dec.com, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
> History has shown that getting bad data in a cache has had major
> headaches associated with it.
Ummm, excuse me, but history has only shown that caches which cannot
know whether or not they have received the whole message, due to the
lack of a content length or chunked encoding, are susceptible to storing
incomplete responses (bad data). What Jeff is talking about is stale
data (data which was known to be good at some prior time, and may simply
be out-of-date).
> I believe we should stay with Jeff's conservative algorithm, and this is why
> I have not to date made any changes in the specification; I will of course
> go with the consensus of the group, but to date, I've not seen
> a good case made by Roy. Not surprising users is a very good thing;
> if the web is unreliable, it will generate many more problems than whatever
> small performance gain might be available with a more optimistic algorithm.
I'm sorry, but this is incredible. I demonstrated how the entire
network of New Zealand (3.6 million people and the highest tech ratio
of any nation) can be completely disabled by Jeff's "conservative algorithm".
Would it be more compelling if I explained how someone could do the same to
Digital's corporate network by spoofing NTP responses to the proxy outside
the firewall?
There is no way that I will implement Jeff's notion of having every
proxy add an Age header -- since I have documented the alternatives,
implementing Jeff's mechanism would subject me to liability every time
some idiot plays network ass-of-the-day.
...Roy T. Fielding
Department of Information & Computer Science (fielding@ics.uci.edu)
University of California, Irvine, CA 92697-3425 fax:+1(714)824-4056
http://www.ics.uci.edu/~fielding/
Received on Thursday, 29 August 1996 16:04:28 UTC