Importing djb's public domain daemontools?

[Warner Losh]

On Jan 17, 2012, at 12:34 AM, Doug Barton wrote:

> On 01/16/2012 23:10, Warner Losh wrote:
>> 
>> On Jan 16, 2012, at 10:10 PM, Doug Barton wrote:
>> 
>>> On 01/16/2012 19:41, Jos Backus wrote:
>>>> On Jan 16, 2012 6:53 PM, "Doug Barton" <dougb at freebsd.org>
>>>> wrote:
>>>>> 
>>>>> On 01/16/2012 12:53, Jos Backus wrote:
>>>>>> 
>>>>>> Thoughts?
>>>>> 
>>>>> This is already available in ports.
>>>> 
>>>> I realize that.
>>> 
>>> Good, then we're done. :)
>> 
>> Not necessarily...
>> 
>>>> If FreeBSD had a solid solution out of the box, all this pidfile
>>>> hackery in the base system wouldn't be necessary.
>>> 
>>> We don't do religious wars here. We especially don't do trollbait
>>> from djb acolytes. The "pidfile hackery" that we currently have
>>> works just fine in the vast majority of cases. The fact that it
>>> doesn't meet some people's ideas of architectural purity is totally
>>> beside the point.
>> 
>> This isn't a religious war.
> 
> You obviously haven't spent a lot of time dealing with djb'ites. Your
> warning sign should have been "messy and unreliable pidfile concept"
> from the OP, or "pidfile hackery" above.

I have spent time with djb-ites in other areas.  I tend to ignore their ranker and focus on the technical issues.  I've had issues with pidfiles and such in the past.  There are a lot of hacks to get around those issues, and things mostly work.  If there's a good alternative that can be demonstrated to work and gain us additional functionality, I'm all for it.  I've fought with init() to make it keep important daemons around should they die.  I've worked with other systems that make it easy to do and miss that on FreeBSD.  It is possible, but not easy.  If daemontools makes it easy, we should evaluate it.

>> This is someone coming to us and saying that it might be a good idea
>> to clean up the mess by importing a tiny bit of extra code
> 
> That's not even close to an accurate description of what this project
> would entail. Have you ever used daemontools?

I haven't.  However, without a fully formed set of patches to test and evaluate technically, it is hard to know if this is a good idea or a bad idea.

>> I'm not convinced it is a non-starter.  I'd fully support Jos if he
>> wanted to commit the code and had done the leg work to do it. 
> 
> One would hope that it would take more than just your support to
> entirely change the way that we start and manage services in FreeBSD.
> 
> Also, see my followup to Jos' subsequent post.

And one would hope your pig-headdednes also doesn't keep it out of FreeBSD.  Neither you nor I are the final arbiter of what's good for FreeBSD.

Warner