Re: Known Problems Draft - Cache-Control absence for Network Transparent Proxies

From: Gary Tomlinson (garyt@novell.com)
Date: Tue Aug 03 1999 - 12:51:32 MDT


Bertold, Ted;

Ted has very eloquently described the relationship of proxy/surrogate/
"http accelerator". We (Novell) have a good deal of experience of deploying reverse
proxies in front of web sites. In these deployments, the reverse proxy is completely
delegated as a surrogate by the web server to represent it as the publishing point,
effectively making the surrogate the origin server.

On the related "oh the irony" topic, I know of several "http acclerators" that can handle large numbers of connections. When deployed as a surrogate, the "http accelerator" provides a simple way to scale up a web site connection capacity.

regards
Gary Tomlinson

>>> <hardie@equinix.com> 08/03/99 10:32AM >>>
Bertold,
        If I understand the question, the answer is yes: clients are
not aware that a reverse proxy/surrogate/"http accelerator" is present
when they send their requests, so they are unlikely to include cache
directives in the request. Theory states, however, that the reverse
proxies/surrogates/accelerators don't need to be controlled with them
since they can be better synchronized with the origin server. I think
that theory derives partly from the origin servers' ability to control
these devices better (since the devices act on behalf of the origin
server) and partly from the historical implementation which sometimes
had the accelerators on the same boxes as the origin servers.
                        regards,
                                Ted Hardie

> Hi,
>
> Is it reasonable to ask about reverse proxies (or accelerators) in this
> case? They would work similarly as transparent caches with respect to the
> client behaviour.
>
> Clients are not aware of using a reverse proxy (or web accelerator) and
> thus they may not send caching related headers which may influence the
> object in the response. In this way reverse proxies can enforce their own
> expiration model instead of the client's own rules.
>
> Bertold
>

> Hi,
>
> Is it reasonable to ask about reverse proxies (or accelerators) in this
> case? They would work similarly as transparent caches with respect to the
> client behaviour.
>
> Clients are not aware of using a reverse proxy (or web accelerator) and
> thus they may not send caching related headers which may influence the
> object in the response. In this way reverse proxies can enforce their own
> expiration model instead of the client's own rules.
>
> Bertold
>



This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:26 MST