[mod_python] Persistent variable again...

I think we're actually in agreement here...

My point was not that the handler is the wrong place for these types of
things, only that the handler is not a core part of the mod_python package.
Having just tackled a new handler and framework, I too would love to see
some more visible framework solutions for mod_python.  That being said, one
of the best things about mod_python is that it isn't really a web app
framework.  It's a barebones implementation that requires some sort of
framework (a handler or more) for an actual app to be developed.

Similarly, I don't think a session handler is 'bad', just that making it a
part of the mod_python package weighs down the implementation (for those
that don't want/use it).  One of the reasons open source development works
so well is that the software is _highly_ componentized, thereby allowing
grisha to focus on his part and whoever to focus on something like a session
handler on top of mod_python (but not as a part of it).

I think we all see value in having the functionality available, and it
appears there's just a disagreement on how it should be made available (in
relation to mod_python).

Just my thoughts.  Take 'em or leave 'em.

Chris

-----Original Message-----
From: Matthew D. Wood [mailto:woodm at equire.com]
Sent: Wednesday, April 18, 2001 2:10 AM
To: mod_python at modpython.org
Subject: Re: [mod_python] Persistent variable again...


I have to disagree.  I think that mod-python (and mod_snake) handlers are
perfect places to put session handlers.  To me, the mod_python framework is
somewhat equivelent to the PHP module only more flexable.  PHP implemented
sessions, and there is no reason why mod_python should be innapropriate for
the
task.  It may be a bit difficult, but not impossible.  If Apache 2.0 has
some
sort of shared memory, then obviously, this is a potential place.  Other
places
certainly include cookies and databases.  The clear advantage of the shared
memory scenario is speed.  This is definitely the case for a
session.data_base_connection variable.  You would have to access the data
base
way to many times just to do a simple query.  I hate having to recreate
database connections every page, and I don't like leaving one open all the
time.  I would much prefer to open it once for each user, then close it when
things time out.

I guess I don't understand why a session handler would be so bad.  If the
goal
of mod_python is to give access, through a python interface, to the
internals
of apache in order to create pages or other environments, why can't we just
treat sessions as a feature, or object that we create in a handler?

I'm not suggesting that mod_python should come packaged with session
capabilities built in.  NO WAY!  What I am saying is that a session handler,
that somebody develops and posts on the mod_python page, could be available
and
tremendously usefull.  I know that mod_snake comes packaged with 3 or 4
"typical" handlers that you can choose to use, or not.  One is a python-cgi
accelerator, one is an embedded python handler; why would a session handler
not
be appropriate?  Then, the users could have a base session handler that they
could modify as they see fit.  If they want a more specific thing, specify.
If
you want a more general thing, either remove some code, or at worst, you
have
an example to work with.

Just my opinions.






Chris Hagner wrote:

> On the issue of supporting modules (i.e. user sessions), there are simply
> too many ways to implement something like that to make it worth coupling
it
> to mod_python.  You'll either try to be too much for too many people (the
> classic framework error) or you'll be so custom to your needs that your
> audience is just you (not a bad thing, just not a place for a solution
like
> mod_python).
>
> I completely agree with Grisha that even the handler is outside of the
core
> mod_python functionality.  That being said, a listing of available
handlers
> on the web site would be very useful for those starting/considering using
> mod_python.  I do think that having this area up there (even it's a bit
> spares initially) will motivate others to share their handlers/frameworks
> that are used with mod_python.
>
> What other handlers have people written?  I've got one that implements a
> servlet-like call mechanism (thereby avoiding the urls being mapped to
> script files).  It's not for everyone, but a perfect example of a decision
> that was made for a specific need.
>
> Chris
>
> -----Original Message-----
> From: Gregory (Grisha) Trubetskoy [mailto:grisha at modpython.org]
> Sent: Tuesday, April 17, 2001 11:25 PM
> To: Damjan
> Cc: mod_python at modpython.org
> Subject: Re: [mod_python] Persistent variable again...
>
> In my (humble) opinion, things like this are outside the scope of
> mod_python, since the goal of mod_python is apache/python integration -
> not building web application frameworks. In other words, mod_python is
> something that makes Apache internals available in Python to give the
> developer more power and speed, it's not something to make web application
> development necessarily easier.
>
> There are a few of application frameworks out there that support
> mod_python - webware is one of them. I don't know much about those things,
> but I think they have features like state keeping, etc.
>
> (To that extent, even the mod_python.publisher handler is outside the
> scope of mod_python, but I am willing to pursue it since it is immediately
> usable, it serves as good test for advanced mod_python capabilities and
> it's a pretty a good mod_python example program.)
>
> Ultimately, there should be an area on the website that lists mod_python
> modules that do various things that aren't part of mod_python
> distribution. Right now there aren't enough of them out there to justify
> it, but eventually there will be I'm sure.
>
> So if you develop something that serves a particular need, please post it
> to the list, and eventually we will have a collection of mod_python
> modules/examples/whatever that people can use.
>
> Grisha
>
> On Wed, 18 Apr 2001, Damjan wrote:
>
> > > I'm trying to implement session like mechanism using mod_python. If
> > > interpreter is created in each processes, is there a way to make a
> > > variable really persistent? I think I should make some daemon that
> > > connects thru named pipe to get/pass session information. Is there
> > > any other way?
> >
> > I've been thinking about session managment for mod_python too, and it
> seems
> > like an important issue. I'd sugest we develop a common interface to
> session
> > managment, and then write specific modules that handle the job, by ways
of
> > cookies, databases, special daemons etc...
> >
> > Any ideas?
> >
> > --
> > Damjan Georgievski            |           Ð?Ð?мÑ~Ð?н
> Ð?Ð?оÑ?гиÐ?Ð?Ñ?Ð?и
> > Skopje, Macedonia             |           Ð?Ð?опÑ~Ð?,
> Ð?Ð?Ð?Ð?Ð?ониÑ~Ð?
> > _______________________________________________
> > Mod_python mailing list
> > Mod_python at modpython.org
> > http://www.modpython.org/mailman/listinfo/mod_python
> >
>
> _______________________________________________
> Mod_python mailing list
> Mod_python at modpython.org
> http://www.modpython.org/mailman/listinfo/mod_python
> _______________________________________________
> Mod_python mailing list
> Mod_python at modpython.org
> http://www.modpython.org/mailman/listinfo/mod_python

_______________________________________________
Mod_python mailing list
Mod_python at modpython.org
http://www.modpython.org/mailman/listinfo/mod_python