I started in 2003 or so, and got into it more seriously because I ended up with customers to support using Perl software. Perl is an amazing language, and with many modern perl approaches, it's wonderful.
This being said like most frameworks it is extremely important to know when to leave the framework or leave things out. While I LOVE working with Moose, my own projects are very heavily stored procedure centric and therefore an ORM like DBIx::Class really does me very little good. DBIx::Class by all reports is an excellent ORM but it's still an ORM. ;-)
There is no better glue language I have ever encountered than Perl We use it in LedgerSMB to glue together stored procedures, web templates (in Template Toolkit), and the web server interfaces. It's a wonderful language and I can't recommend it highly enough.
As for Perl 6, it has joined HURD 1.0 as one of those projects that will be released when pigs fly. I suspect that Perl 5.x will be as far as Perl gets.
David Wheeler (theory) started working on a stored procedure centric project a while back and as a result extracted the DBIx::Class connection logic into DBIx::Connector - which I have a feeling you're already using.
I spent a fair amount of time hanging out in #ledgersmb a while back and in spite of being the project founder of DBIx::Class, I don't think I ever felt I had a convincing argument for using it, so honestly I'm glad you're not doing so (if I ever come up with an O<sproc>M or something I'm sure you'll find out :)
Actually some of your recommendations are pending getting rid of old code (we simply can't move everything to PSGI when we have the mixture of old and new code we have, so it's CGI only for now still). However we are giving hard thought as to the best way to do so and I expect that as we continue that process things will become more PSGI friendly until it is just a matter of changing one sub somewhere.
I should note you have convinced Josh Drake on the merits of DBIx::Class for other projects though.
Finally I think it is worth noting that I think that in the 1.4 tree we've come out (I hope) of the contageous effect of the bad SL code, and so the coding styles which evolved so much during LSMB 1.3 are being formalized and moved to Moose in 1.4.
This being said like most frameworks it is extremely important to know when to leave the framework or leave things out. While I LOVE working with Moose, my own projects are very heavily stored procedure centric and therefore an ORM like DBIx::Class really does me very little good. DBIx::Class by all reports is an excellent ORM but it's still an ORM. ;-)
There is no better glue language I have ever encountered than Perl We use it in LedgerSMB to glue together stored procedures, web templates (in Template Toolkit), and the web server interfaces. It's a wonderful language and I can't recommend it highly enough.
As for Perl 6, it has joined HURD 1.0 as one of those projects that will be released when pigs fly. I suspect that Perl 5.x will be as far as Perl gets.