Re: ICP & HEAD requests

From: Chris Wedgwood <chris@dont-contact.us>
Date: Mon, 7 Sep 1998 13:15:17 +1200

On Sun, Sep 06, 1998 at 11:48:48PM +0200, Henrik Nordstrom wrote:

> Do we even need to differentiate between HEAD and GET in ICP? What
> are the chances that a cache only has a cached HEAD reply?

Zero. I can't make squid cache HEAD objects, only GETs.

> Perhaps sending normal ICP queries on HEAD requests is enought?
> Squid does answer HEAD requests when it has a "fresh" object in
> cache, and I hope most other caches does the same...

Yes, HEAD requests are satisfied when fresh objects are in the cache.

> Hmm... while playing around with HEAD.. is HEAD headers cached at
> any time? I think I have seen evidence of this in the code, but I
> can't get either my 1.2beta or 1.1.21 to act as it does.. I only
> get HEAD hits if there is a fully cached object (from a GET
> request).

1.2beta24 does not appear to cache HEAD objects. Using a clean cache,
I was unable to make HEAD requests come from cache until I actually
issue a GET, after that time, HEAD request are satisfied from cache.

Normally I would argue that I don't think make sense - HEADs don't
occur very often, and are fairly small. Its hard to justify the
ability to cache the head only... except, we could consider a HEAD to
be a partial-request of zero bytes, and we do now support
byte-ranges, so in effect HEAD request could become cacheable and for
the most part would make this argument somewhat moot.

-Chris
Received on Tue Jul 29 2003 - 13:15:53 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:11:54 MST