[OpenSIPS-Users] auth_db module in 3.2.2
liviu at opensips.org
Thu Aug 12 16:04:42 EST 2021
On 12.08.2021 18:36, Adrian Georgescu wrote:
> The auth_db module has some dramatic changes which are either
> undocumented or not backwards compatible and is unclear how to handle
Indeed, with the addition of RFC 8760 support (support for SHA-256 and
SHA-512-256 auth algorithms), me and Maksym Sobolyev decided to try and
remove the "ha1b" feature, originally designed to accommodate some
broken SIP UAs who cannot follow the basic SIP authentication spec. The
feature had been in there since the very beginnings, and we were not
sure if anyone is really benefiting from it anymore nowadays.
A strong reason for removing "ha1b" was the sheer number of hashes to be
stored per subscriber. Since we now have 3 algorithms (MD5, SHA-256,
SHA-512-256), there are 3 hash-columns to store. With the "ha1b"
feature, there would be 2 x 3 = 6 hashes in total to store, per user.
So you can see where this is going: /"Can we get away with dropping ha1b
and storing half the data per user?"/ ... was the big question.
Still, we agreed that if there is still enough traction for the "ha1b"
feature from the community, we can easily re-add the ha1b logic and 3
more columns to the table and backport everything to 3.2. It's a
trivial task, frankly.
The big question is: on your platform(s), can you control the software
in all SIP UAs that incorrectly include "realm" information in the
"username" field (which should really be just the *user**'s **name*!)
and fix the problem on the phone side?
PS: I noticed the 3.2 migration page is missing any info on ha1b. Will
get it fixed soon, depending on the outcome of the discussion.
www.twitter.com/liviuchircu | www.opensips-solutions.com
OpenSIPS Summit 2021 Distributed | www.opensips.org/events
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Users