[31789] in North American Network Operators' Group
Re: decreased caching efficiency?
daemon@ATHENA.MIT.EDU (Adrian Chadd)
Fri Oct 20 12:52:39 2000
Date: Sat, 21 Oct 2000 00:47:05 +0800
From: Adrian Chadd <adrian@creative.net.au>
To: Dana Hudes <dhudes@hudes.org>
Cc: nanog@merit.edu
Message-ID: <20001021004704.R94031@ewok.creative.net.au>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <00c201c03aa6$cc786bc0$3d5cdcd1@hudes.org>; from dhudes@hudes.org on Fri, Oct 20, 2000 at 11:02:42AM -0400
Errors-To: owner-nanog-outgoing@merit.edu
On Fri, Oct 20, 2000, Dana Hudes wrote:
> > >When I switch to CGI-based delivery of images the cache will of course
> > >become pass-through
> > >since there will be no file to cache just a stream of bytes....
> >
> > Is the assumption there that by using CGI you'll automatically tweak a
> > configuration in a caching proxy? If so then it's a flawed assumption.
> >
> But there is no file to cache? I don't have enough gear to set up a test with squid myself
> (and that would only be one cache) but how is the engine to know to cache it?
> My understanding is that CGI-generated content is usually not cached.
BZZT. Another assumption which is actually totally not true.
For example, imagine your photo book. The photos won't change, right ?
The position in your database won't change, right ? So ..
http://www.domain.com/photos?id=31765
ok. 31765 is a static image that won't change. So, you'd be better off
setting its expiry time to something high, wouldn't you?
> > Having had a very quick look at your site, it seems a little strange that
> > you want to defeat caching of those objects that soak up bandwidth; the
> > request to perform "click-through" on the advert suggests that you're using
> > the revenue to pay for your bandwidth costs. (So, one assumes that the
> > more the material was cached, the less you'd have to pay, and the less
> > you'd have to worry about page impressions.) I particularly like the way
> > that you require my browser to send a Referer field to be allowed to view
> > the pictures ;-)
>
> I do indeed use the revenue to pay for bandwidth but the pictures, by and large
> (its a work in progress) have been tuned for file size; still takes time to decompress but hey,
> what can I do. Also the projected load vs. the bandwidth is such that I have a LOT more room left. The users get a reasaonbly large bitmap in a reasonably small file. ImageMagick is nifty set of programs. The problem I have is pirates who collect images and use them for other purposes.
> the pictures...well, I actually don't want them hanging around on the user's disk once the browser is no longer on the page.
> I haven't figured out how to make that happen other than expiration of 1 minute or something.
You can't. End of story. This is the internet, people control their end-nodes,
so you have zero chance of this happening. If you *REALLY* want to be evil,
you wrap the images in a java applet so they can't just rightclick on it,
but again that won't stop the smart people.
Adrian
--
Adrian Chadd The Law of Software Development and
<adrian@creative.net.au> Envelopment at MIT:
"Every program in development at MIT expands until it can read mail."
