RE: Peer twiddling

> > > > Right now the SNMP code puts its fingers into the squid code.
> > > > I don't think thats right.
> > >
> > > I agree, but with the caveat that we don't provide any
> alternative! We
> > > have to export the squid data in a machine format
> somehow, and we don't
> > > even do that internally! (The stats pages are _html_ out!?!)
> >
> > I don't. I like SNMP for machine reading, and I don't see SNMP stick
> > it's fingers more into the Squid code than the cache_object class of
> > functions, only differently.
> >
> > The stats pages are raw text, not HTML. cachemgr.cgi
> translates it to
> > HTML using very simple rules. But the thing with the stat
> pages is that
> > they are intended to be read by a human, not a computer.
>
> *nod* So should they be formatted to be read by a computer, and have
> some external program/programs handle converting them to other
> formats?

Well, we could just change them to use XML, and then use one of the
N-thousand
XML parsers available.

> What about comments on the original idea? :-) The guy who I'm doing it
> for would like it to appear in -HEAD and then 2.4 so he
> doesn't have to
> run a "custom squid" (I like people like this. :-)

I have no strong feeling wrt this.
The interface could be made a bit more resilient to human errors (I don't
like
having the option to shut down squid happily intermingled with inspector
methods in the cachemgr) maybe, but as I said it's a matter of interface,
it could very well be delegated to the external translator (be it cachemgr
or something else).

-- 
	/kinkie