sql server replication: I've written a custom conflict resolver in the form of a stored procedure and
applied it as the resolver for a particular article in my merge replication.
This stored procedure is failing because the subscriber is not in sysservers.
I use the 4 part name to look at subscriber data like the sample code from
MS.

Do I have to manually setup a linked server
once I set my stored procedure as a custom resolver? There was nothing in
the documentation that pointed this out and I assumed that once a subscriber
connected and instantiated the replication that it would be good to go.

We have hundreds of subscribers so if this is the case what are the
ramifications of setting up hundreds of linked servers?

I suppose I should clarify the situation.

We're using merge replication where the subscribers are on tablet PCs
running MSDE and they are not part of the windows domain. The merge process
is kicked off on the tablet PC by an application using the SQLMerge component.

I've written a custom conflict resolver as a stored procedure. This works
as long at the tablet PC is sitting on the network. The @subscriber is
passed into my stored procedure as "machinename\instancename" and I can
establish a linked server programmatically inside the procedure. Then I can
query the subscriber information using the 4 part name and the rowguid.

Will SQL Server give me an IP address in place of the machine name for the
@subscriber if the tablet PC initiates a merge over the internet?
"Machinename\instancename" will work for a linked server over the internet,
I'll need "ipaddress\instancename".

[quoted text, click to view]