[asterisk-users] Discussion: Are we ready to leave 1.4 behind?

[Administrator TOOTAI]

Le 28/04/2011 22:43, Leif Madsen a écrit :
> On 11-04-28 04:33 PM, Administrator TOOTAI wrote:
>> Le 28/04/2011 21:47, Leif Madsen a écrit :
>>> On 11-04-28 12:04 PM, Administrator TOOTAI wrote:
>>>> Ok, so why not stay with asterisk 1.4 security *and* bug/regression fixes for
>>>> few weeks/monthes till 1.8 reaches the level that the community accept to switch
>>>> to 1.8
>>> What is the guide here? What is the "level that the community" accepts?
>>> Unfortunately that is a statement that is impossible to measure quantitatively.
>>> The answer will always be, "We're not ready!"
>> Don't think so, analyze the answers to this discussion -thanks Ole ;-)-: till
>> 1.8 is not at the feature level and stability of 1.4, people like me will not
>> move to 1.8 Measure is easy :-)
> But that's what I don't get. No one is *forcing* you to move to 1.8 *right now*.
> The code base for 1.4 isn't going anywhere. Anyone is able to keep deploying 1.4
> (or 1.2, or 1.0, or 0.9 for that matter) to their hearts content.

Sure. Please follow the 2 next stories:

- had a customer running 1.4.26 We upgraded to a new server and 
installed 1.4.39, last version at this time. Bang: voicemail doesn't 
work as it should, had to fallback to 1.4.26 Customer is still running 
this version.
- have 1.4.41 and 1.6.16 which are no more able to use auth keys in iax 
since we update one server from 1.4 to 1.6

Now imagine that 1.4 stays at only security level. For first case we 
have 2 options: upgrading for security reasons to last version but then 
no more voicemail, or staying with 1.4.26. In the second case, upgrading 
both servers to test with 1.8. If it's still not working, it was time 
loose beside other problems.

Yes, we have servers for testing, but really, who would think that such 
2 problems araised with an 1.4 stable version? Same was few versions 
before (1.4.20~1.4.28 if I good remember) with attempted call transfer: 
was working on one version, stop to next one, worked again aso. Even in 
a test environment you can't simulate all setups.

Hope that this both scenario gives you a new vision ;-) and why I tell 
that bugs and regressions should be taken in account at the same level 
as security.

-- 
Daniel