[geeklog-users] Avoid hardcoding http protocol in {site_url}

[Drago Goricanec]

Actually, the protocol and the base are two different issues. I added the base
so that I wouldn't have to rewrite image URLs that are referenced by the story
(they're stored in the database with the protocol). Also, I wanted links to my
local site to match the protocol that was currently in use. They could all be
dynamically rewritten. Perhaps rather than storing the full URL, there could be
a [site_url] prefix that was stored in the link or article and rewritten when
the pages are generated. It would also make it easier to move the site to
another URL.

The protocol stuff was as clean as adding this line to my config.php:

$_CONF['site_url'] = ($_SERVER['SERVER_PORT'] == 443 ? 'https:' : 'http:') .
"//www.example.com/";

What is the method of submitting patches to the development team?

Drago

Quoting Tony Bibbs <tony at tonybibbs.com>:

> I personally wouldn't like this because I know some proxy servers that 
> pass web traffic to web servers behind it end up making the relative 
> pathing a problem.  Geeklog needs to make clear (and clean) support of 
> SSL for all admin functions.  The base is there we just need to change 
> the code to allow for https for admin stuff.  It maybe as simple as 
> adding a $_CONF['admin_protocol'] variable that can be either http or 
> https.  Then we could dynamically generate the https for the admin side. 
>   We should also make a similar change to the public login page.
> 
> Oh, hope the proxy server thing makes sense. We have a configuration 
> here at work like that and it reaks havoc if you don't use absolute 
> paths (nor just for our PHP apps but our Java and ASP/.NET apps as well.
> 
> --Tony
> 
> Drago Goricanec wrote:
> > I modified geeklog code to detect whether http or https was used to
> connect, and
> > then write all URLs using the resulting protocol. The reason is that I
> didn't
> > want to leave either my password, or my weblog information on a computer
> or
> > network that I didn't trust (when administering the site), but I didn't
> want
> > all users to be forced to use https.
> > 
> > Current code uses $_CONF['site_url'] for all URL references. If 'http:' is
> > hardcoded here, then even if the connection is over 'https', all references
> to
> > images and links within the site remain 'http'.
> > 
> > I changed all URL references to my geeklog site to be relative to
> {site_url}
> > (i.e., {site_url} doesn't appear in any of the templates). I changed the
> layout
> > header to emit <base href="{site_url}/">. {site_url} still exists, but it
> is
> > dynamically created based on the protocol used to connect (http or https)
> with
> > a path to the weblog. Trailing '/' is omitted. All other URLs, such as
> > {site_admin_url} are relative to {site_url}. Even images in stories are
> stored
> > with a relative URL. 
> > 
> > I also added a secure/normal login link to switch between http and https
> just
> > below the login form.
> > 
> > Would others be interested in this change? I can provide a patch against
> the
> > latest source. What is the process of submitting such patches?
> > 
> > Drago
> > _______________________________________________
> > geeklog-users mailing list
> > geeklog-users at lists.geeklog.net
> > http://lists.geeklog.net/listinfo/geeklog-users
> 
> 
> _______________________________________________
> geeklog-users mailing list
> geeklog-users at lists.geeklog.net
> http://lists.geeklog.net/listinfo/geeklog-users
>