Daniel Vainsencher <danielv at netvision.net.il> wrote:
> If we make updates use SM to update in-image packages, we must rely on
> map contents. But we need to make sure that it is up to date, and that
> we get not errors in the process.

Right.
 
> BTW, updates are a relatively very good time to do loadUpdates, since
> one is quite likely to be online when running them (someone running
> updates from disk can probably figure out whats going on and fix it).

True.

> A related issue - I've sworn to myself that I would not request any more
> SM features until 1.1 is out. But I'm weak, so here goes anyway. Anyway,
> though important, I think this feature should wait for 1.1.1.
> 
> I think it is important that SM not bring up spurious exceptions when
> doing stuff, especially if we're going to rely on it to do various kinds
> of updates. Therefore SM should implement the following two invariants -
> * If it is connected to the net, it should bring up no spurious
> exceptions, because it does loadUpdates and retries the failing action
> before raising one.

True. I could try to catch Exceptions generally and do something logical
about them.

> * If not connected to the net it can't know the real state of the global
> map, but it should act as if some version of the map it has is
> definitive. Either the one represented by the log files, or the one
> represented in the image (and which dominates should be made clear). It
> should never complain that things are out of synch, because the user
> simply doesn't care.

This was actually very much remedied in the latest SM 1.0x. It now
simply picks the one which is most up to date. It never complains
anymore.

> Daniel

regards, Göran