← Back to team overview

launchpad-dev team mailing list archive

Re: generic job queue cronjob?

 

On Monday 27 September 2010 16:07:06 Aaron Bentley wrote:
> I am sorry that using the Job API has been a problem for you.  I would
> appreciate feedback on how we can improve it.  I don't want you having
> to bend it.

No need to apologise - the Job API is fine itself, the problem is that we 
already had a different API for the build farm which is quite different to the 
Job API.  Various people wanted me to try and make the build farm API work 
like a Job, which ended up being responsbile for race conditions, data 
corruption and reduced follicle counts.  I wish I'd pushed back more and kept 
our existing API, because I still don't see its advantages over the old one.

Because of this, I'm really sceptical that the experience will be better when 
making the other Soyuz APIs work in the same manner, so the fewer moving parts 
the better.

> >, and partly because modelling everything as a job is not useful
> >
> > when all we want is a simple event.
> 
> We will have to examine this use case in greater detail at some point,
> because so far, I don't understand why you wouldn't want a Job.

See above :-)

I have absolutely no objection to talking about it and I'm very happy that 
you're taking the lead on this.  I just don't want you to close any doors too 
early.

J



References