ZXTM 5.1 released

Its good to see more options are being introduced in how cached content can be managed. The low-inpact cache refresh is great, but what I'd like to see is some form of stale-while-revalidate and stale-if-error option.

http://www.mnot.net/blog/2007/12/12/stale

We at the BBC would really value this!

Ross,

The new cache-refresh settings give you something that is almost identical to the stale-while-revalidate option in the article you cited.

Let me explain... suppose you wish to cache content for 120 seconds. Once the cached version expires (is stale), you want to continue to serve the cached version for a further 60 seconds. If anyone requests the content during that 60 second period, you update it (and reset the cache timeouts); if no-one requests it, you drop it. That's the 'stale-while-revalidate' behavior as I understand it.

So, in ZXTM, set the cache time to 180 seconds, and the Virtual Server's webcache!refresh_time setting to 60 seconds. In the last 60 seconds of the cache time, ZXTM will continue to serve responses from the cache, but it will send a maximum of one request per second to the back-end server, and update the cache when it gets a response.

This gives identical behavior to the stale-while-revalidate option, with one exception; the cache is not updated asynchronously, and a maximum of one request per second is forwarded to the back-end server until the cache is updated (so may incur a delay).

Section 18.2.2 of the 5.1 manual describes this in more detail.

On the other hand, there's not an easy way to emulate the stale-if-error option, but we've already got a feature request open for it. Thanks for raising it as a suggestion.