[Gnumed-devel] client v vs database v mismatch and write-protecting old databases

[James Busser]

Like Geordie, I am in a situation where I have GNUmed v9 database  
when my client is only 0.2.8.10-1 and I wanted to revisit what  
currently happens, and what we desire to happen.

It seems to me that there is a risk of allowing a client which was  
built for a prior database to be able to connect to a newer version  
of the database. My concern is that even if the client may appear to  
work OK meaning it does not throw errors, it can still happen that  
the new database makes possible the entry of more information which  
the user does not realize. The result can be missing information.

Now in the other direction, if a client *would* make it possible to  
lookup something from an earlier version of the database (maybe  
because the new server is down, or there is a suspicion of altered  
values that do not make sense to the clinician) then I would be OK  
with the same client being able look back at the old database, not  
that it would be a guarantee that it would work, however if it did  
work, then I cannot see that it would be unsafe, especially if we are  
talking about reading vs writing. The old database could.should have  
had its precautionary notice row inserted. There is a risk (no matter  
which client someone is using) that the ability to look in the old  
database creates a risk that they will write to the old database.

What would be the easiest way of making an old database readable but  
not writable? Is it anything configurable at the level of postgres?