Strange title, I know. But hear me out.
I talked before about creating easy to deploy "AOLserver
Applications". This is do-able the way I described it, but to do it
right, quite invasive in the core. And it probably still won't be
simple enough for people used to PHP deployments.
Here are some other thoughts to make deploying a new app as simple as
it is in PHP:
- On demand pools
I looked into the nsdb code and nothing seems to indicate that you
can't define pools at any time you please. So a special "ns_db
gethandle" that takes something like "nsdb:postgres:localhost:5432/
mydb?user=foo&password=bar&timeout=10&max=20" should be
implementable. It would create a handle name from the arguments
(rewrite them; people could use a different order of arguments in
different places...) and then see it the pool exists. If not, define
the pool and then continue as normal.
All this would require from the user is editting the main config file
with:
ns_section "ns/db/drivers"
ns_param postgres nspostgres.so
in fact, "base.tcl" could do a "file exists" for every know database
driver and load it if it is there.
- On demand libraries
create an "ns_libload" command. sources a tcl file from the given
location (relative), in the global scope. In a special namespace in
the interpreter, an array is kept of which have been loaded and what
the mtime was. With the "ns_libload_mode" command you can set it to
"always", "check" and "persist". Check would be default; ever time
the ns_modload for that file is called, the mtime is checked; if it
is new, load it. This would overwrite any procs with the same name
that already exist. Because every interpreter loads any library only
once, this should be just as efficient as the current libs. (when set
to persist, of course!)
Good practice would be to only ever source 1 library in any .adp,
which in turn takes care of sourcing anything, but that is up to the
user.
- .nsaccess files
Work pretty similar to Apache's .htaccess. They are sourced by
"walking back" the directory tree and the first one found counts for
all subdirectories. These could have a section of stuff that needs to
be executed before the actual URL is processed. This means you could
do a request to /news/story/1234.html and the .htaccess in the news
directory would be called first and does an ns_registerproc to handle
"/news/story*". A run-once section could be used to schedule procedures.
The last on is the most invasive in the core, but the first two seem
very do-able. The library thing can be done entirely in Tcl, probably
in an hour or two! But it's no good without also doing the first.
(database on demand)
My goal with this all is that all a user needs to do to install
"ns_wiki" is to:
# cd /usr/local/aolserver/servers/server1/pages
# tar -xvzf ns_wiki.tar
# mv ns_wiki wiki
# chown nsd wiki/wiki.cfg
A "setup.adp" could ask for database detail, work out where it is so
it can register the correct urls and take all other sorts of config
paramters.
Of course, none if this will change the behaviour of any existing
AOLserver apps, which is the beauty of this solution!
Any thoughts?
Bas.
--
AOLserver - http://www.aolserver.com/
To Remove yourself from this list, simply send an email to <[EMAIL PROTECTED]>
with the
body of "SIGNOFF AOLSERVER" in the email message. You can leave the Subject:
field of your email blank.
