Re: Efficient way of enumerating squid contents?

From: Chris Wedgwood <chris@dont-contact.us>
Date: Mon, 12 Oct 1998 12:11:00 +1300

On Mon, Oct 12, 1998 at 12:38:56AM +0200, Henrik Nordstrom wrote:

> Well.. you have to be very sure there. Microsoft-IIS/3.0 is not a
> guarantee that the URL-path is case-insensitive. There may be a
> module (or whatever they call it) installed that handles the URL
> and not a normal filesystem. Not that this is widely used but it is
> a possibility.
>
> It is also remotely possible that a cgi-bin, asp or similar
> "program" is sensitive, even thought the server in general is not.

Both of these are indeed true - and probably not as uncommon as you
think. I think we stuff here that does both, certainly, some URLs are
case dependant for some things, but most are not.

But - that said, for example, I think if the following _all_ apply,
then case folded is a pretty safe bet:

  - the URL ends in .asp
  - the server is Microsoft-IIS/3.0
  - the reply has a Last-Modified date
  - no cookies, content negotiation or authentication are used

I originally only brought this up because I was bitten by this. My
first thought was the redirector interface, but it seems inadequate.
As for how much difference this will make, at the suggestion of
Stepher Baker I checked...

from approx. 250,000 URLs :

    - 89046 unique requests, no case folding

    - 88811 unique requests, case folding

So, assuming nothing special about the size distribution, and
assuming these differences are indeed for identical objects (two big
assumptions), then the savings by implementing this would be
negligible (less than 1%), but nonetheless, if I want to implement
this, or some variant of it, I am unable to do so, because the
redirector interface is too limited.

-cw
Received on Tue Jul 29 2003 - 13:15:54 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:11:56 MST