Hi all, i am about to release a new uWSGI wizard, but before committing it
i would like to hear some comments or objections :)

Currently uWSGI supports 3 configuration file format:

XML
INI
YAML

and 8 "magic" extensions that will setup the server with standard
options

.wsgi (will load a .wsgi script)
.py (will load a python script, not as a module)
.psgi (will load a psgi script)
.pl (will load a perl script)
.ru (will load a rackup script)
.rb (will load a ruby script)
.lua (will load a Lua script)
.ws (will load a Lua wsapi script)


So i think it would be better to reduce the uWSGI wizard at a simple 
text input field that will take one of this file.

The code will read the file extensions and if it is one of the 3 configuration
files, it will parse them searching for the "socket" option and (eventually) 
the mountpoints.

If "socket" is not specified the wizard will bind to an automagically choose a 
port (as now).

Automagically bind will happen for magic extensions (in addition -M option will 
be added)

Notes:

YAML parsing requires the python yaml module that could be not available in the 
system, so i will
insert a check that will not show .yml extension in the form if it s not 
available.

Should uWSGI logging be disabled by default if not specified in config files ?

Without config files uWSGI will spawn a single worker. What do you think about 
setting
the number of processes based on the number of cpus in the system ? (a ncpu*2 
should be a good default value)

The newer uWSGI codebase contains a new option "manage-script-name" that will 
manage suburl-application mapping
itself (without the help of the webserver). I think it could be better to 
disable mountpoint parsing in case it is specified and simply
"mount" the uWSGI source under /

Thanks for your attention
--
Roberto De Ioris
http://unbit.it

_______________________________________________
Cherokee mailing list
[email protected]
http://lists.octality.com/listinfo/cherokee

Reply via email to