Thanks Christopher for clearing it up!

and thanks Marcin for taking the time to check on it.

Maybe this should be a bit more clearly stated in the documentation?

It can sort of be read between the lines but clearly stating it would 
not hurt :)

cheers
Martin


Christopher Browne wrote:
> Martin Eriksson <m.eriksson at albourne.com> writes:
>   
>> hmm could be i've missunderstood something...
>>
>> but it appears that move set does not change all the subscriptions to
>> the orgin node?
>>
>> It appear that if i got
>> 1 -> 3
>> 1 -> 2
>>
>> and i do move set
>> from 1-> 3
>>
>> 1 will subscribe from 3, but 2 will continue to subscribe from 1.
>>
>> how should I go about moving subscription on node 2 to point to 3 instead?
>>     
>
> You can submit a SUBSCRIBE SET request to point node #2 to use node
> #3.
>
> Slony-I doesn't assume that *everything* should get redirected; it
> tends to do the minimal change (which is, after all, what you asked
> for!).  This is, in effect, the difference between DWIS and DWIM, that
> is, Do What I Say versus Do What I Mean.
>
> Historically, systems that try to implement DWIM tend to be rather
> less predictable than those implementing DWIS, because their behaviour
> has rather more complex policy bundled into it.
>