httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marc Slemko <ma...@worldgate.com>
Subject Re: The trouble with benchmarks
Date Wed, 15 Apr 1998 23:54:56 GMT
On Wed, 15 Apr 1998, Dean Gaudet wrote:

> Sure there is, see www.mirror-image.com.  Just because rfc2068 doesn't
> give it an explicit name doesn't mean it's non-compliant.  It's just a
> proxy cache whose existence is transparent to the client. 

But it isn't truly transparent.  You can define it as being a transparent
cache, sure.  But the concept itself is doomed to failure for broad
definitions of transparent.

My definition of transparent is that it doesn't change the way the request
or response appears to the server or the client.

The typical definition of transparent just says that the client doesn't
have to explicitly know about it. 

A slightly more complex definition involves the client appearing to come
from the right IP; Vixie's boxes do this so that the source address of the
request going through the proxy cache is actually the address of the
client, but this still isn't transparent and can, in fact, be very
dangerous.  The problem is that HTTP does not handle varying based on IP
address with proxies well.  Make one client make a request that is allowed
based on the origin IP address.  Then when another client using the same
proxy makes a request, it is cached so they get the content even though
they shouldn't.  I do not call that transparent.

Note that this isn't a problem with a normal proxy that uses its address
to source requests because _all_ clients can get the same documents
through the proxy.

It all hinges on definitions, and I don't buy into the definition of a
transparent proxy because it causes too many side-effects however you do
it to be called transparent.  

> 
> Dean
> 
> On Wed, 15 Apr 1998, Marc Slemko wrote:
> 
> > On Wed, 15 Apr 1998, Dean Gaudet wrote:
> > 
> > > Why should it identify itself?  It's a transparent proxy.  The client
> > > doesn't even know of its presence. 
> > 
> > There is no such thing as a transparent caching proxy.
> > 
> > Transparent proxy, sure, but they aren't much good for too many things
> > other than firewalling.
> > 
> > > 
> > > Dean
> > > 
> > > On Wed, 15 Apr 1998, Ian Kallen wrote:
> > > 
> > > > Do you mean PV's caching thingy (http://www.mirror-image.com/)?? It does
> > > > _not_ I identify itself as a cache?  Bleh!
> > > > 
> > > > At 02:47 PM 4/15/98 -0700, Dean Gaudet wrote:
> > > > >And don't go saying "but 90% of the clients are win95" ... because
that's
> > > > >a false path to wander down.  In addition to traditional proxy caches,
> > > > >there are transparent proxies that sit between clients and the webservers
> > > > >which do no identification of themselves.  See www.vix.com for example.

> > > > 
> > > > 
> > > > --
> > > > Ian Kallen						ian@gamespot.com	
> > > > 	Director of Technology and Web Administration
> > > > 		GameSpot Incorporated
> > > > http://www.gamespot.com/		http://www.videogames.com/
> > > > 
> > > 
> > 
> > 
> 


Mime
View raw message