8 Apr
2009
8 Apr
'09
1:52 p.m.
On Sun, Apr 05, 2009 at 04:05:59PM +0200, Holger Freyther wrote:
Hey Harald,
I didn't know your favorite way of getting patches so for now I simply
attached them to this mail. I found some time to continue with channel
management and would like to get feedback if this is running into the right
direction or if you have something else or other ideas.
I think it looks great, please continue this way. Obvious bugfixes like the
call state one should be directly committed to svn.
The reason to not just apply these patches are within
the first patch.
Currently we don't serialize the struct gsm_bts so after a restart of the
bsc_hack we would have to force another location updating request...
I think we can do something like 'lazy re-establishment', i.e. if we get a
channel request and identify the subscriber, then we should just treat this
as a 'imsi attach' and assign the current_bts pointer.
Also, the attach/detach state (and number of bts as well as timestamp) has to
be stored in the sqlite database, so we after bsc_hack restart we can do a db
lookup and decide if we should try to page (and where!) or not.
At bsc_hack startup we can iterate over the database and expire any old
'attach' state entries if they are too old (user-configurable timeout, but
I'd
say one hour as default)
regards.
--
- Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)