[asterisk-dev] About asterisk development plans

[Tzafrir Cohen]

On Wed, Mar 18, 2009 at 07:58:19AM +0100, Olle E. Johansson wrote:
> 
> 17 mar 2009 kl. 20.32 skrev Jon Bonilla (Manwe):

> > Many people I know are going back from lastest 1.4.2X releases to  
> > 1.4.14-1.4.17

1.4.17? You mean the one before the SIP security advisory?

> > because it seems that newer releases are more inestable than older  
> > ones. After
> > two years of work in a frozen branch I think it should be rock-solid  
> > and changelogs should be minimal but they are not.

My problem with such a sentence is that it is "hearsay". Those versions
fixed quite a few bugs that bothered me and overall left Asterisk more
stable. Pointing to specific regressions would be more useful.

>
> It is indeed a big problem that I meet all the time. Do you have any  
> proposal on how to change it?

So this is what happens when you try to backport too many bug fixes to a
2 years old branch. And the backporting is even mostly done by the team
that wrote the code in the first place.

> I personally think that the new release policy makes 1.6 impossible to  
> use in any production environment. I've said that many times, so it's not  
> any news for any one. I'm maintaining a private version of Asterisk 1.4 
> for my carrier customers with backports of smaller changes in the 1.6 
> branch (all available in my svn directory). The cost of evaluating a new 
> version is high and 1.6 release policy is not acceptable, since the core 
> keep changing too fast.

Let's face it. Features are coming in fast. You use a software like
Asterisk because it is a wildcard. With the old model you had to wait
for over a year to get the latest developed feature in a "supported"
version. This means that:

1. Those new features were not tested well enough
2. You often had to run an unsupported version in production
3. If your personal stabilization cycle missed Digium's one, you get an
extra delay of up to a year in getting new features.

Another thing to keep in mind is that we're not exactly the Linux
kernel. The Linux kernel model is highly decentrlized: there is, indeed,
the kernel.org kernel. But it is mainly used as a focus point, and not
supported in the long-run. Most users get the supported versions from
Linux distrubutions.

> 
> The plan I've outlined earlier on this list, which got support from  
> Russell,
> was at some point to fork the 1.6 code and create a kind-of "Centos"
> distribution with long-term support for this group. So far, no  
> customer has
> shown interest and with the state of the SIP channel in the 1.6 tree
> (Digium doesn't like me saying that, but anyway...) I don't see
> it coming soon. For such a branch to succeed, we will need a test
> team that works with it and a developer team that focuses on it,
> which requires external funding. Large scale carrier installations is  
> not
> within Digium's commercial focus today.

I already see some distributions use Asterisk 1.6.0 . People already use
1.6.0 in production. At some point in the future 1.6.3 will be relweased
and 1.6.0 will no longer be supported.

At this point those users will be faced with the choice of either
upgrade or find a different source of support. I believe that when such
a decision point will come closer, you'll find such an idea more
popular.

-- 
               Tzafrir Cohen
icq#16849755              jabber:tzafrir.cohen at xorcom.com
+972-50-7952406           mailto:tzafrir.cohen at xorcom.com
http://www.xorcom.com  iax:guest at local.xorcom.com/tzafrir