Re: cachemgr interface

On Wed, Mar 03, 1999 at 01:41:09PM -0700, Alex Rousskov wrote:

> The original (not mine) idea was to have Squid output HTML-free and
> do all the formatting in cachemgr.cgi or other scripts. This
> approach may be a good compromise between
> very-human_friendly-and-hard-to-script_parse output and
> very-human_unfriendly-and-easy-to-script_parse output.

Since it will almost always be used by cachemgr.cgi or similar, I
think we want as-easy-to-parse-as-possible. I don't know of anyone
who uses this directly (eg. via telnet)

> An alternative is to mimic cool interfaces some commercial products
> have. We will probably be looking at writing stuff in Java then.

We can do both -- if the format is clean, we can write a java applet
to do the above

> A more complex question is how much pre-computing (not formatting)
> of raw values should be done in Squid. For example, Squid currently
> maintains many histograms and other data structures that are not
> really needed for Squid operation. They are required for some
> administrative tasks though.

IMO squid should maintain the bare minimum is has too, which I
beleive is pretty much the case

> It is not clear if Squid should spit out raw data only and rely on
> external scripts to accumulate and process that data. If Squid
> should do internal preprocessing, it is not clear what the limits
> for such preprocessing should be.

Squid should do as little as possible. I think what it does now is
sufficiently simple

-cw
Received on Tue Jul 29 2003 - 13:15:57 MDT