Re: "Hits" pragma

Brian Behlendorf (
Sun, 13 Aug 1995 14:24:01 -0700 (PDT)

On Sun, 13 Aug 1995, Koen Holtman wrote:
> Brian Behlendorf:
> >[..Statistics to be sent by proxy caches..]
> >So, it looks like a structure of
> >
> >host timestamp referer
> >
> >would satisfy a most applications.
> Make that
> URI timestamp host_or_just_domain_prefix [referer] [from] [request-id] .

You don't need the URI in the request structure because the request is
linked to the URI itself, and sent when the URI is re-requested.

> The [referer], [from], [request-id] fields are all optional, they
> should be included by the cache if the user agent request to the cache
> included them.

Right, I wasn't suggesting the proxy invent values for referer if it
didn't exist :)

I would certainly be happy with From and RequestID if they were a part
of the structure, but I was trying to find a balance between no data and
every HTTP header imaginable.

> In my opinion, domains that want to use their proxies to get a high
> level of privacy should be allowed to do so.

I totally envision starting up! :)

> If all domains decide they want maximum privacy, we get 'open
> rebellion' again (though I hate to use the term `rebellion' for
> something initiated by the markering department), so maybe there has
> to be some incentive to give away privacy. Just shipping popular
> proxy cache and browser software with statistics-friendly default
> settings might be sufficient.

Right. Or as I pointed out in the last message, the servers could be
smart and give more "compliant" caches longer Expires times as a reward
for their diligence.

> To draw a parallel: I don't see the Usenet community inventing
> protocol extensions to make spamming use less IP packets, they just
> try to educate companies about spamming being a bad idea.
> The web community can do the same thing. Maybe we should begin by
> inventing an interesting term for the practice of service providers
> disabling caching just to get better statistics.
> cache piracy?
> marketing packets?
> Orwellian URL?
> In addition to making protocol extensions, we can try saving the world
> through the Jargon Watch section of Wired. I am not kidding: this is
> a serious, though postmodern, proposal.

Violent agreement - this is really subtle subject right now for most
people, but perhaps we can fight marketing with marketing :) Emphasize
the concept that high cacheability improves reliability and response time
and hits-per-server-dollar, and maybe they'll listen. A "Hits" pragma
will really help this.


--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-- http://www.[hyperreal,organic].com/