Right now, we rely on Django's caching for performance. We use the django-
staticgenerator app to do some ad-hoc extra caching.
There are some pages which are accessed by users who are logged-in that we should
be able to cache. Right now, if a request gets passed all the way to Django, we
have Django generate the results.
But that is not the only way to do it. Instead, Django could simply indicate that
the request has not been modified in the time since it was last cached.
The way this would work is:
* First time: Client connects to our transparent proxy and does a GET. Proxy has
no data in the cache, so it asks the Django code. Our Django code returns the
full response, as well as this HTTP header: "Cache-control: public, s-maxage=0,
proxy-revalidate".
* Second time: Client connects to our transprent proxy and does a GET. Proxy has
the page in its cache. Because the Django app said 'proxy-revalidate', the proxy
must always ask the server if the cached version is valid. The proxy does the
GET, but also adds a header: "If-Modified-Since:" plus the time that it did the
GET that caused the data to be in the cache. Django app looks at all the models
that it would have used, and if none were modified since the If-Modified-Since
time, it says, "304 Not Modified". This is way faster than actually generating
the page.
Thanks to Michael Stone for this concept.
|
