Re: [Cgi-session-user] RT#224 and CGI::Session::PureSQL

[Mark S. <ma...@su...>]

Ron Savage wrote:
> Hi Mark
> 
> Examining CGI::Session::PureSQL I see it starts with:
> use base qw(
>     CGI::Session
>     CGI::Session::ID::MD5
> 	CGI::Session::Serialize::SQLAbstract
> );
> but there is no package CGI::Session::ID::MD5, but there is a
> CGI::Session::ID::md5 (lc MD5). So, CGI::Session::PureSQL is not really
> production-ready, right :-)?

Hmm, I used it production.

> Also, I had to delete t/01_serialize_sql_abstract.t to make progress.
> 
> And what's this stuff about Indiana time? Even here in Australia we know
> there's no one dead or alive /really/ living in Indiana, right :-)))?

It turned out tests gave different results depending on what timezone 
they were run in, which was very frustrating to debug. (They passed for 
me, but not for some other people once they hit CPAN). If I had to 
revisit that again, I would try one of the newer Perl development 
modules that fake what the time is.

> Should I work on CGI::Session::PureSQL (hint: No) or should I pursue the
> idea of getting the DBI-oriented code to accept specific names for
> columns, not just for MySQL, but Pg, and (gasp) my own code:
> o CGI::Session::Driver::odbc
> o CGI::Session::Driver::oracle
> 
> I can see I might add $IdName and $DataName say alongside $TableName in
> CGI::Session::Driver::DBI and to have that package's descendents also
> have local vars of the same name and to default the values from the
> parent. That way, users could set the 2 col names the same way they can
> set the table name. What do you think?

First, it would be interesting to know that the module you found was for 
CGI::Session 3.x.

The 4.x version was eventually completed and published here:

http://search.cpan.org/~markstos/CGI-Session-Driver-pure_sql-0.62/lib/CGI/Session/Driver/pure_sql.pm

Because the SQL it uses is very simple, it is likely to already work 
with several DB backends.

    Mark