Fallout from the CVS discussion

[Warner Losh]

On Sep 16, 2012, at 6:34 AM, Eitan Adler wrote:

> On 16 September 2012 01:35, Konstantin Belousov <kostikbel at gmail.com> wrote:
>> On Sat, Sep 15, 2012 at 04:37:49PM -0400, Eitan Adler wrote:
>>> However, -CURRENT is not meant to be a production system.
>> 
>> It is simply not true.
> 
> My statement was true, but does not disagree with the content below.
> Production system != Production Grade.

One of the things we are trying to move towards is that current can be cut into a release branch on short notice.  We need to keep it as close to production ready as possible.  People do put -current systems into production for testing purposes, or because they have made the evaluation and know what they are doing.  Discouraging production systems from current, the present project policy, doesn't mean the project doesn't appreciate the people that do put current systems into production and the data that generates.

Put another way: saying that current isn't meant for production systems as a justification to slash things out before we are quite ready isn't something people in general want to encourage, since it is close to attitudes in the past that got us into a lot of trouble.  Sure, in this case the reaction is a bit of hyperbole, but there's long, historically lingering wounds that put people on a hair trigger.

>> CURRENT shall never be knowingly put into a state
>> where it cannot be used for the 'production-grade' use, whatever it
>> means.
> 
> Agreed.
> 
>> We do accept changes are so disruptive that some unknown fallout
>> is expected, since otherwise developers cannot make any significant
>> progress.
> 
> The point of my statement is that it perfectly acceptable to change
> behavior in HEAD in a non-backwards compatible way.

Some behaviors, yes.  Most behaviors need to remain the same for a variety of reasons.

> In particular no
> systems running -CURRENT are expected to be "critical functioning".

Yes, they often are.

> People that track -HEAD are expected to be able to deal with the sorts
> of problems that occur from "drastic change."

Generally yes.  However, we do try to cushion the blows that are delivered in -current.  The reason we have the separation isn't so we can do whatever we want in -current, it is so that when somebody messes up, the damage is more limited.

>> But introducing known breakage is simply not acceptable. Doing so shrinks
>> the already limited testing base we have for HEAD.
> 
> Agreed.

Ditto.

Sorry to be so pedantic on pushing the point in this meta-discussion, but I don't want to see us slide back into the 'wild west' that current was in the 5-current time frame.  The CVS thing, by itself, wouldn't do that, but we must have the proper attitudes for getting change done, and when we pull the trigger on change.

Warner