On Wed, Apr 3, 2013 at 1:16 PM, Andreas Gal <[email protected]> wrote:
>
> On Apr 3, 2013, at 10:07 PM, Jonas Sicking <[email protected]> wrote:
>
>> On Wed, Apr 3, 2013 at 1:01 PM, Milan Sreckovic <[email protected]> 
>> wrote:
>>> Agreeing with everybody, even though there is a bit of a disagreement :-)  
>>> There is the "right thing" and the "thing we have to do".  I know it's hard 
>>> for me to talk about "things we have to do" because, well, that leads to 
>>> imperfect world.
>>>
>>> The reason the trains work for us is because we have everybody looking at 
>>> nightly, so by the time we start slowing down for Aurora and Beta, we know 
>>> we're in decent shape, and we know we can change our behaviour, lock things 
>>> down, and hit the dates for shipping.
>>>
>>> With the current setup, we have a large group of people, with a large set 
>>> of tests, get on the train late and introduce the work that we're not used 
>>> to doing at that stage of the game.  To beat the analogy to death, it seems 
>>> like we should have another station, or another set of tracks for them.  In 
>>> the perfect world though, they would be on the train from day one, and not 
>>> surprise us and themselves with all the work way late in the game.
>>>
>>> I'm getting to the point - if our partners can get on the aurora train 
>>> right now (22), then maybe there is enough time to ship based on that 
>>> version.  I doubt our partners will go for it, but it would be something to 
>>> start planting now.  As for riding the 23 train, I like to think I'm brave, 
>>> but that one scares the crap out of me, with a few working days between us 
>>> producing a build and pencils down.
>>
>> As far as I know, we have shipped 16 releases of Firefox using the
>> train model without ever being more than a few days late. So I'm not
>> worried about having the non-b2g parts of gecko ready for the target
>> date here. There's a lot more risk in the B2G work, but that risk is
>> unaffected by which train we are on.
>>
>> In fact, optimizing for reducing that risk means that we should do
>> whatever makes b2g development easiest. And the shorted distance that
>> we have between m-c and the train we are shipping, the easier writing
>> b2g code is.
>
> You are asking for different things here. You are proposing two things.
>
> 1. Switch from 18 to 22 for v1.1, which means taking hundreds or thousands of 
> patches that were never tested or certified as part of v1.0, dramatically 
> changing our risk profile. This is a complete no-go for our partners. We can 
> keep arguing about this, but it will absolutely not happen. You have the 
> vendor contacts yourself. Feel free to reach out and ask.

Yup. This was one of the points that I listed in the original emails.
I don't want to speak for Alex, but I *think* that he has offered to
do this.

FWIW, I think the large number of patches that we've taken for B2G
since certification poses a much larger risk than the gecko patches
that we are talking about here. Simply because the B2G patches are
much closer to the relevant code.

/ Jonas
_______________________________________________
dev-b2g mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-b2g

Reply via email to