> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Stephen Burke
>
> Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Daniela Bauer said:
> > If you implement a feature/bug somebody is going to use it/exploit
> it/fall for it, whether on purpose or accidentally.
> > I can't believe "we are relying on the users to deal with this" still
> counts as an argument in this day and age.
>
> In the case of glexec, "users" means "VO framework developers", so it
> isn't so unreasonable to expect them to get it right.
>
It's still the wrong approach. It should be done properly because it's
the right thing to do. It doesn't matter how few or competent we think
the users are - shipping something that's buggy and expecting it to
be worked around is just not on.
Besides which, it sounds to me like some of this stuff, particularly
what to do about working directories, is site configuration anyway,
so the VOs aren't in the best place to deal with it.
> We basically only have three clients which will use glexec, i.e.
> the analysis pilot jobs for atlas, lhcb and cms (alice will do
> it differently it seems).
>
I'd expect the minor VOs to follow in due course. We've already
been discussing getting them using a framework (probably Dirac)
rather than direct WMS based submission. I know that if I was
looking after a small VO I'd want to be using infrastructure that
was as similar to the LHC VOs as possible because I'm less likely
to hit bugs, corner cases, site oddities and other forms of general
weirdness that no-one cares about.
Ewan
|