http://bugs.meego.com/show_bug.cgi?id=2254

pohly <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |syncevolution-default-bugs@
                   |                            |meego.bugs
          Component|SyncEvolution               |SyncEvolution
            Version|1.0                         |unspecified
         AssignedTo|[email protected]      |[email protected]
                   |                            |gs
            Product|OS Middleware               |SyncEvolution
           Severity|normal                      |enhancement

--- Comment #1 from pohly <[email protected]> 2010-05-18 05:59:44 PDT ---
(In reply to comment #0)
> Besides the "personal" covered by the HTTP http server UI submitted, there is
> also a usecase for those having a small home server. In this scenario, it's 
> not
> really easy to use the http server as of today.

I agree, this could be simplified. My own thinking was to remove the need for
separate D-Bus startup script and roll that into the syncevo-http-server.py
itself, including starting of syncevo-dbus-server and setting env variables
when installed in a non-standard location.

> My own thoughts on this are that with two reasonable changes this should work
> quite fine.
> 
> The first is to allow the http server to run without Seahorse. Seahorse is
> really not useful in this scenario, and adds a lot of dependencies including 
> an
> accessible X display (even if it's not used).

If the username/password are set inside the config.ini files, then the GNOME
keyring should never be accessed. Is that failing for you?

> For the headless case maybe just
> validating the user's  login password would be enough.

Unfortunately there's no way for a user space program to access the password.
Access to a plain text password is required by the SyncML protocol.
Alternatively, someone could implement HTTP auth (bug #1349) and then disable
password checking at the SyncML level.

> The other fix is to allow the server to run from inetd. This would give
> advantages w r t server load and http server stability (since it's just
> restarted for each sync). This means that it should be able to use 
> stdin/stdout
> instead of socket communications when given a command line option.

That looks worthwhile, but the implementation is a bit more complex: the
syncevo-http-server<->syncevo-dbus-server communication is not stateless. The
syncevo-http-server must remain running throughout a SyncML session and handle
all POST requests possibly related to it. A client which cuts the connection
between messages would end up starting different instances of
syncevo-http-server, if I'm not mistaken.

-- 
Configure bugmail: http://bugs.meego.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.
You are watching someone on the CC list of the bug.
_______________________________________________
Syncevolution-issues mailing list
[email protected]
http://lists.syncevolution.org/listinfo/syncevolution-issues

Reply via email to