cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vadim Gritsenko <>
Subject Re: Releasing from trunk
Date Wed, 07 Feb 2007 13:31:54 GMT
Carsten Ziegeler wrote:
> Andrew Savory wrote:
>> Hi,
>> Not wishing to spread FUD and rain on the parade, but ... I think 2.2  
>> is *massively* less tested than 2.1.x. The main reason 2.1.x goes out  
>> with not much testing is because it's been used in production by a  
>> very large number of people. How many are actually using 2.2 in  
>> production right now? Carsten, Daniel, Giacomo, Reinhard ... I guess  
>> at most < 10 sites?
>> I'd love to see 2.2 out the door and more people using it, but I'd  
>> also hate to see people's first experience of 2.2 be a buggy one.
>> Given that we're still seeing changes enough to kill Cruisecontrol,  
>> I'd suggest an 'M' is more suitable than an 'RC'.


>> And yes, I'm putting my money where my mouth is and trying to use 2.2  
>> to build a site ;-)
> Great :)
> We need to get a 2.2 out; imho it is more important to have something
> out after so many years than having a 115% bug free version.

You are missing a point. It's not about a bug free version. It is about a 
version which seen more than like 4 hours of testing after major surgery.

> In addition, there is no guarantee that releasing a milestone will bring
> us more users testing, more documentation etc. And I personally doubt
> that people will start trying a milestone release.


They are not even trying, they are using it in production. It is a bit different 
for 2.2 though: there were no "real" downloadable release yet, so there is 
really no easy way to download 2.2 and give it a spin on existing 2.1 project.

High entry point would hamper wider adoption of 2.2, not the label on its release.

> And what will be the exit criteria? When do we have enough confidence
> (if it is really missing today) that we are stable enough for switching
> from milestone to rc? We only have indicators and feelings. I think that
> the current code base is stable and the work we are doing with is a
> proof for this (for me).

When there is a consensus in the community that version is ready for RC or Final 
release. Given the responses so far, you already can see that to reach consensus 
we'd have to:

   * have some more testing done (not 4h or 1d as now),
   * have unit tests working,
   * have CC working and not failing over,

> On the other hand, a RC does not mean bug free - it means stable from
> the api and implementation pov. If we would be sure that it's bug free
> we could release a final right way.


> Yes, we had several changes in the last days/weeks, but doing a RC gives
> ourselves the responsibility to not change these things until we have a
> final version out (or create a branch).

Do you mean to say you can't stop doing major refactoring and start polishing 
unless it bears RC badge? What happened to good old-fashioned discipline and 
self control? :)


> So, in short: a RC is an indication to *everyone* that this is the a
> stable version from the api pov and that we think (or hope if you like)
> that this version is stable as well. It's more likely that people will
> try out an RC than a milestone. So if we get feedback back at all, this
> will happen for an RC but never for a milestone.
> Carsten

View raw message