perfsonar development work

[Uros Juvan <>]

What about requiring LS registration to be implemented in the service, but
which can then be disabled.

If I remember correctly, you get errors in the error log if service.r.ls_url
property in service.properties is empty. Maybe we should just improve this,
so LS Registrator scheduler is not run if that property is missing/empty?

On Tue, Oct 10, 2006 at 09:17:43AM +0200, Maciej Glowiak wrote:
> Loukik Kudarimoti napisał(a):
> >Maciej,
> >
> >LS registration process is simple and a registration makes data 
> >discovery easy. I have no doubt about this. I wouldn't like to force 
> >this onto a developer. I would like the developer to see the advantage 
> >in the process and then put the feature into their software. A perfSONAR 
> >service definition should include minimum set of requirements and for me 
> >this only includes schema compatibility (i.e. NMWG v2 schemas)
> >
> >Nobody forces a business to advertise themselves into Yellow pages. 
> >Businesses see the advantage and do so. If a service doesn't have the 
> >feature but the user would like to register with the LS, the user would 
> >request the developer to put that feature in the next release.
> >
> >All this said, I am happy either way. I would prefer the way I mention 
> >above but I am quite happy to compromise.
> >
> 
> 
> Hi Loukik,
> 
> I understand your point of view (but I don't share it :)
> I think registration is very important thing and surely will distinct
> perfSONAR services from non-perfSONAR applications that use only NMWG
> protocol.
> 
> But perhaps there will be some cases when registration is not necessary
> or even difficult to achieve (but I am not talking about Roman's case
> because I think the service he mentioned should have registration option).
> So maybe we should say: all pS services should have registration
> capability or there are some important reasons not to implement it.
> 
> It would prevent us from "validating" services that don't implement
> registration only by developers' lazyness.
> 
> Maciej
> 
> 
> >Loukik.
> >
> >Maciej Glowiak wrote:
> >>Roman Lapacz napisał(a):
> >>>
> >>>I see your point. But assume that I want to create a simple python MP 
> >>>which stores data in existing Java RRD MA. If I did it without the 
> >>>registration feature (because I decided that it's useless for me and 
> >>>I didn't want to put more efforts) then couldn't I call it a 
> >>>perfsonar service?
> >>
> >>In such definition -- yes.
> >>I don't think it's so compicated to create registration. It's just 
> >>sending one request to the service.
> >>
> >>Think about people who just use MP you mentioned. Some of them may 
> >>want to register such MP to LS and what then? That's perfSONAR, so we 
> >>must agree for small set of common functionality. I think such 
> >>functionality is LS registration.
> >>
> >>Maciej
> >
> >
> 

Regards,
-- 
Uroš Juvan
Arnes