What's left before 2.0, and what's not in CVS.
What's left before 2.0 is released? There has been little feedback about the -pre releases (and -pre2 was unusable for some things I tried). What's not in CVS, and what I will commit shortly: * removal of HUP support. It doesn't work. * move almost all logic to time/fd-based event loop (like libevent, but simpler). This means radiusd.c becomes tiny: start.. while (do_events() == HUP) { // handle HUP } * framework for HUP support is in place. Maybe we can start adding HUP in 2.0.1, etc. * virtual servers for pools of home servers. e.g. a pool can point to a virtual server, just like clients or listeners. The pre-proxy and post-proxy methods for a request are run from that virtual server, rather than the ones in the "client" or "listener" sections. This means that auth/acct packets can be handled by different pre/post proxy methods, which is nice. * rename of "server_pool" to "home_server_pool", as there are too many references to "server" everywhere. * New xlat's: %{client:...} %{home_server:...}, and %{home_server_pool:...} The goal is to enable per-client/home server/whatever policies. Put "foo=bar" into a client, and you can do: if ("%{client:foo}" == "bar") { ... } This enables policies to be per-client, without adding more configuration files to the server. I think that's it. The only thing I'd *like* to add is HUP support. But realistically, there isn't time. And it's WAY past time to release 2.0. Thoughts? Alan DeKok.
Alan DeKok wrote:
What's left before 2.0 is released? There has been little feedback about the -pre releases (and -pre2 was unusable for some things I tried).
What's not in CVS, and what I will commit shortly:
* removal of HUP support. It doesn't work.
* move almost all logic to time/fd-based event loop (like libevent, but simpler). This means radiusd.c becomes tiny: start.. while (do_events() == HUP) { // handle HUP }
* framework for HUP support is in place. Maybe we can start adding HUP in 2.0.1, etc.
* virtual servers for pools of home servers. e.g. a pool can point to a virtual server, just like clients or listeners. The pre-proxy and post-proxy methods for a request are run from that virtual server, rather than the ones in the "client" or "listener" sections. This means that auth/acct packets can be handled by different pre/post proxy methods, which is nice.
* rename of "server_pool" to "home_server_pool", as there are too many references to "server" everywhere.
* New xlat's: %{client:...} %{home_server:...}, and %{home_server_pool:...} The goal is to enable per-client/home server/whatever policies. Put "foo=bar" into a client, and you can do: if ("%{client:foo}" == "bar") { ... } This enables policies to be per-client, without adding more configuration files to the server.
I think that's it. The only thing I'd *like* to add is HUP support. But realistically, there isn't time. And it's WAY past time to release 2.0.
Thoughts?
We still need to get the memory leak issue sorted on Darwin. Sorry I haven't been able to provide more feedback.. I'll have another look at it today and see if it I can narrow I down. Also need to decide on handling of reply attributes inserted pre-proxy.
Alan DeKok. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/devel.html
On Tue 09 Oct 2007, Alan DeKok wrote:
What's left before 2.0 is released? There has been little feedback about the -pre releases (and -pre2 was unusable for some things I tried).
While I have been peripherally following development, I have been busy with other things recently and haven't had the cycles to do significant testing... -- Peter Nixon http://peternixon.net/
On Tue, Oct 09, 2007 at 10:49:04AM +0200, Alan DeKok wrote:
What's left before 2.0 is released? There has been little feedback about the -pre releases (and -pre2 was unusable for some things I tried).
There are a couple outstanding rlm_ldap bugs I've fixed in our local copy, but haven't made it into the tree. - #483 fixes a ldap xlat bug which chops off a character from the result. - #458 + #459 fix compare_check_items. Thanks, -- Chris Mikkelson | Another couple of days in this place, and I'd start cmikk@qwest.net | spouting new age psychobabble, talk about sitting on | top of a mountain contemplating a tangerine for the | rest of my life. -- Tony Bourdain
Chris Mikkelson wrote:
There are a couple outstanding rlm_ldap bugs I've fixed in our local copy, but haven't made it into the tree.
- #483 fixes a ldap xlat bug which chops off a character from the result.
Fixed.
- #458 + #459 fix compare_check_items.
Wild. It looks like no one really used that code before. I'll take a look at committing the fix before 2.0-final. Alan DeKok.
Hi, the mySQL schema patch? And will there be another -pre? Or maybe called -beta, to give more people the confidence to try it out? Stefan -- Stefan WINTER Stiftung RESTENA - Réseau Téléinformatique de l'Education Nationale et de la Recherche Ingenieur Forschung & Entwicklung 6, rue Richard Coudenhove-Kalergi L-1359 Luxembourg E-Mail: stefan.winter@restena.lu Tel.: +352 424409-1 http://www.restena.lu Fax: +352 422473
Hello, i know, this is an unusual way to submit a patch, but i'm not used to use the normal bug system. This one is very small and it enables alternatives in recursive calls. Example: %{module:call(%{A2:-default})} The current code breaks the main attribute at the :- and ends with an error. (I guess, the 'int spaces' variable was introduced to avoid this behaviour in SQL statements). The patch was made against the latest CVS-Snapshot (20071024). Sorry for not using the normal way, M. Ruttmann
Matthias Ruttmann wrote:
This one is very small and it enables alternatives in recursive calls.
Example: %{module:call(%{A2:-default})}
The code already supports this. See "man unlang", and "conditional syntax". What you're missing is that your proposed fix was in 1.x, AND was intentionally deprecated in 2.0. Why? It doesn't work in the general case. The %{%{foo}:-%{bar}} syntax is better and safer. Alan DeKok.
participants (6)
-
Alan DeKok -
Arran Cudbard-Bell -
Chris Mikkelson -
Matthias Ruttmann -
Peter Nixon -
Stefan Winter