I have wrote the following text, assuming the approach would be that the PHP.net webmasters decide of the indexing options, so all what the mirror hosters need to do is retrieve the configuration files from www.php.net,
and just change the spidering address to their own mirror. Would be happy to hear suggestions/ideas... :)
This looks good as a start. The hu.php.net maintainer [Cece] also plays with finetuning the indexing options for us, and building some better interface for mnogo...
We need to decide whether we would like all searching mirrors to have all languages indexed. That means approx 1GB of index as tested by Cece. That also affects the charsets selection of course.
9. Download the php.net configuration files:
% wget http://www.php.net/mnogosearch/indexer.conf
% wget http://www.php.net/mnogosearch/searchd.conf
% wget http://www.php.net/mnogosearch/search.htm
% wget http://www.php.net/mnogosearch/stopwords.conf
Those will be in mirrorkit.tgz in the mnogosearch dir ;)
10. Now edit the file indexer.conf and find the "Server" directive. Change it from http://www.php.net/ to your mirror's address,
and change the file path to your mirror's DocumentRoot.
We need to pick some dumb URL instead, and not www.php.net, so those forgetting this step will not bog down the server with requests.
12. Edit your Apache's configuration file (httpd.conf).
Find the PHP.net's mirror VirtualHost definition. Append to it the following:
Alias /search/ /usr/local/mnogosearch/bin
AddHandler cgi-script .cgi
13. Restart your Apache:
% apachectl stop
% apachectl start
14. Now try http://yourmirror.php.net/search/search.cgi :)
AFAIK it was decided that we should use some PHP interface to the CGI. Maybe in the form of running it from search.php as a CGI, or by using the PHP extension. [Or offering both methods for mirrors to choose]. I am definitely against using the default mnogo search.cgi interface. PHP has very good support for mnogo, and we will need to finetune several things [eg. default to the users preferred language, etc].
So once again, this is a very good instruction set for a start, but there are some weak points, which we need to discuss and further develop.
Goba