You are viewing a plain text version of this content. The canonical link for it is here.
Posted to rivet-dev@tcl.apache.org by Karl Lehenbauer <ka...@sc.com> on 2004/05/30 19:26:13 UTC
to clean up or not and knowing what might need to be cleaned up
I have been an advocate for Rivet basically not cleaning up anything
other than the request namespace and leaving it up to the developer
(me, etc) to decide what they want to leave around or not.
But as the NeoWebScript guy, my thing was to cleanup everything. In
fact, we did a per-page interpreter and deleted that interpreter after
every page. Kind of the opposite mentality, but it was principally for
an ISP environment where the users were untrusted.
Ironically, there's a bug in DIO or possibly in a big of legacy code we
have that doesn't use DIO that I'm leaking Postgres results, so I'm
actually having the very problem people are talking about. (I haven't
looked hard for the problem.)
OK, how about if we 1) add introspection to find out what has possibly
been leaked? Possibly some standard webpages that'll let you list all
of the globals, the names of the namespaces, the names of all the
packages that are loaded... stuff like that.
And maybe there should be a way to ask it to clean up or to globally to
the application tell it some patterns of things we don't want laying
around.
Really in my case I don't want the thing cleaning up dangling Postrgres
results, although I wouldn't mind it telling me that they're there.
It's a bug I want to track down and fix, not something I want to sweep
under the rug with an end-of-page automatic cleanup. Although the
whole notion of postgres results is a little gross. 95% of the time or
more, the getting of a new postgres result, you're done with the
previous one, so the paradigm ought to be more oriented to that as the
normal case, although that's another conversation for a different time.