HTTP Caching Basics

Friday, May 16, 2014 4:40:24 PM

HTTP Caching Overview

HTTP caches all files types; it has no intrinsic notion of types.
Caching uses two mechanisms:
- expiration. eliminates the need to send requests
- validation. eliminates the need to send full responses
Caching should preserve semantic transparency.
- The client should receive exactly the same response from the cache…
- as it would have received from the origin server…
- had the cache not been involved.
Caching applies the most restrictive (i.e. most semantically transparent) interpretation of conflicting cache-control mechanisms.

General Caching Timeline

The client makes a request to the origin server.

The cache intercepts the request. If the cache contains a cached response, it checks whether the cached response is fresh. If the cached response is fresh, the cache returns it to the client. Otherwise, the cache sends a validation request to the origin server.

The server receives the validation request. If the cached response is valid, the server returns a 304 not modified response. Otherwise, the server returns a first-hand response.

In either case, the cache receives the server’s response. If it’s a 304 not modified, the cache returns its cached response to the client. Otherwise, it caches the first-hand response from the server, then forwards it on to the client.

Caching Techniques

Here are some use cases and caching techniques that will make them happen.

Table: I would like the client cache to…

Purpose	Headers	Notes
Allow caching, force revalidation.	Cache-Control: private, max-age=0	google.com does this.
?	Cache-Control: private, no-cache, no-store, must-revalidate	facebook.com does this.
?	Cache-control: no-cache, no-store	plus.google.com does this.
?	Cache-Control: private, max-age=0, must-revalidate	github.com does this.

HowTo: Inspect Headers with Fiddler

Open Fiddler.
Open the Composer tab.
Type a GET address (e.g. http://www.google.com, http://www.bigfont.ca, etc)
Turn on capture (F12).
Click Execute.
The Inspectors tab will open.
Click Headers.
This lets you view cache related headers.

Fiddler Shortcuts

Use F12 to toggle capture.
Use Ctrl + X to clear captures.

The Cache Related Headers

Common headers / directives.

Headers

	In Request	In Response
Age	n/a	The sender’s estimate of the time since the origin server generated the response
Authorization	n/a	special rules for use in shared caches
Location	n/a	n/a
Pragma	implementation specific (e.g. IE 9)	same
Vary	determines whether to use a fresh response without revalidation	indicates by which fields the response varies
Warning	n/a	warn of lack of semantic transparency or entity transformations
Date	date/time of message generation	n/a
Expires	n/a	date/time after which response is stale

Cache-Control Directives

	In Request	In Response
public	n/a	okay to store in any cache
private	n/a	do not store in a public cache
no-cache	forces revalidation	okay to store but must revalidate
no-store	generally to protect privacy; cache MUST NOT store any part of the request nor any response to it	same
s-maxage	n/a	overrides max-age & Expires in shared caches
max-age	willing to accept ages up to max-age; a zero value forces revalidation	sets the age at which entities become stale; a zero value forces revalidation on each request
min-fresh	?	n/a
max-stale	willing to accept ages beyond stale	n/a
only-if-cached	only return a cached response	n/a
must-revalidate	n/a	never use stale cached entries
proxy-revalidate	n/a	public caches must revalidate; private one do not need to
no-transform	do not change the media type	same

Conditional Headers

Use with methods (e.g. PUT) to make the method conditional. “Server, perform the method if and only if…”

	In Request	In Response
If-Match	an ETag matches	n/a
If-Modified-Since	the entity has been modified since	n/a
If-None-Match	no ETags match	n/a
If-Range	the entity is unchanged, the send me missing parts only	n/a
If-Unmodified-Since	the entity has not been modified since	n/a

Muddling Thru