Gzip compression - pre-compression vs. on-the-fly

[Igor Sysoev]

On Sun, Apr 13, 2008 at 12:01:31AM -0700, mike wrote:

> Just wondering why there is not a standard method of storing the
> gzipped file in a cache directory in nginx. Seems to me that would be
> developed before the pre-compression module.
> 
> Like in Lighty, there is a /var/tmp/lighttpd/cache/compress/ type
> folder, and the server maintains its own cache there, as opposed to it
> being up to the user to manually gzip their files in-place (and the
> pre-compression module just checks if a {$file}.gz exists)
> 
> 1) pre-compression - available in 0.6.24
> 2) on-the-fly compression - standard
> 3) standard server maintained one-time compression - ??
> 
> There could be a good reason for this, and I can see how #1 is pretty
> neat (you control what is cached, don't have N copies of the same file
> gzipped on each webserver's temp cache dir, etc...) but I am surprised
> #3 is not offered.
> 
> I know #2 seems to meet most people's needs and from what I've read
> has minimal overhead. Wondering why #3 was skipped altogether and #1
> was done instead?

As to overhead of #2, I've tried to minimize it, but anyway on my production
servers gzipping takes about 30% nginx CPU time according to google-perftools.

I'm developing cache infrastructure in nginx. The first client of it
will be proxy module. But eventually I will add cache capabilities to other
modules including gzip filter.


-- 
Igor Sysoev
http://sysoev.ru/en/