Re: [Gnumed-devel] what about commit?

[Karsten Hilbert]

[snip]
> Just in order to enforce transactional safety (in the case some smartass 
> accesses a connection directly instead of going via gmPG) we have built in 
> the "safety lock" of the pseudo user name for this.
That was a good explanation. I'll just repeat the detail that
PostgreSQL 7.4 supports read-only connections (sessions) and
that our code already supports taking advantage of that. We
just need to switch it on once 7.4 is widely available (and
released, actually :-)  At that point our workaround with
_$user will be superfluous and can be removed. Which should
not need any changes other than in the SQL scripts and gmPG.
Certainly not in well-written client code.

BTW, since gmPG caches ro connections we don't need to worry
all too much about requesting such a connection on every
gmPG.run_ro_query() invocation.

Karsten
-- 
GPG key ID E4071346 @ wwwkeys.pgp.net
E167 67FD A291 2BEA 73BD  4537 78B9 A9F9 E407 1346