Subject: Re: [vserver] Lets sponsor some development of vserver
From: Herbert Poetzl <herbert@13thfloor.at>
Date: Sat, 23 Aug 2008 13:13:00 +0200

On Sat, Aug 23, 2008 at 11:34:58AM +0100, Ed W wrote:
> OK, several folks have stepped up and offered some funds to buy some 
> contracting time from Herbert - since he also appears to be at a 
> temporary impasse to do continue development it seems like a good time 
> to try and crystalise that.
> 
> I think to make this workable the spec needs to be simple, but also I 
> have to ask Herbert to be considerate to a situation which is poorly 
> defined and not easy for businesses to get their head around.
> 
> I propose that the spec is to "advance development of the next stable 
> vserver release, with the goal of a "released" patch suitable for 
> production use against kernel 2.56.26 or newer". 

there are some open issues (mostly missing features)
with 2.6.26 Linux-VServer which would make sense to
specify here:

 - the Token Bucket Scheduler wasn't ported yet
   (need adaptation to the CFS)

 - Hard CPU Limits are missing (again related to
   the CFS runqueue changes)

 - PID spaces require special handling for init-less
   guests (which is not available in mainline)

 - the switch to PID spaces will break backwards
   compatibility, unless we add some compatibility
   layer (migth be interesting for folks who want
   to stay with older tools :)

 - kernel patch breakdown (split patches) and
   review, to ensure that the 2.6.26+ port is
   complete

those are areas where work has to be done in the
kernel, where I think I can put some numbers on
further we have a few testing areas which definitely
need some love to get to a stable release

 - cross compile tests for more unusual archs
   and/or configs need to be done

 - mainline network namespaces need to be tested
   and finally integrated 


> I would request that Herbert offer a private "timesheet" to those     
> sponsoring the effort to show good faith that the money has been      
> spent as requested. It might also be sensible to request daily        
> checkins or patches to some repo again to satisfy any management      
> types that a real product was produced (details TBA). I hope also     
> that Herbert will offer us a sensible daily rate for this work since  
> I suspect it's all off unless we can get a sensible amount of work 
> done.  

all those things above are quite easy to verify
(even without deep knowledge of the mechanisms)

> The sponsorship will be paid 50% up front and 50% in arrears or some
> other arrangement agreed privately with Herbert
> 
> Can I ask all those who previously offered funds to get in touch with
> Herbert directly and arrange what is reasonable. Herbert can I ask you
> to coordinate this and "make it work" if you are interested in the
> job?

yep, will try to coordinate where possible

> I will pledge 1,000 EU to the pot straight away and depending how the
> pot looks later I may top that up with an additional sum. Can I ask
> everyone else to dig deep and see what we can pony up

> Herbert can you please respond with a credible target that we can 
> achieve.  Please understand that many folks will not contribute except 
> to something with a reasonable fixed price goal and an easily observable 
> target.  So I hope you can describe a goal more concrete than simply 
> "will spend 2 weeks hacking" !


probably the first 'goal' is to get the PID stuff
done, and the second 'milestone' could be the 
scheduler rework, but I'm open to suggestions how
to priorize that ...

> I will pick some arbitrary dates and suggest that we "try and wrap this 
> up by the end of next week".  Herbert to come back privately to those 
> interested with a schedule of work and some estimated times
> 
> Can we make this work?

I hope so, but only time can tell ...

in any case, thanks for the efford!

best,
Herbert

> Regards
> 
> Ed Wildgoose
> http://www.mailasail.com
> tel: +44 20 70432832