Unit testing... I think you missed what the real problem with django is about, or just your projects don't demand to use that kind of features etc.
How about something simple like pretending there we have django 1.0 (not 1.2 which is quite new), and let us try to connect to 3 databases at once with django's ORM?
Oh, we are screwed it can't do that. Ok, then lets try to swap out our ORM to sqlalchemy.... crap all nice plugins broke, admin panel doesn't work etc. This is why REALLY django lack quality and maturity - it's components are not versatile enough - and in my book this is a very serious issue because if my requirements are diffrent from what django makers assumed then I'm doomed. This is why people mention other (in my opinion superior, but not that easy to use), solutions to the problem. My point is - majority of people fail to notice such serious flaws like unability to use more than 1DB at once (yes i know it got finally fixed after few years). I'd rather use one tool that should cover my needs than be forced to switch them because they don't fit the needs of project.
If your application required connection to three databases concurrently, then you almost certainly wouldn't have chosen Django 1.0 as your application development platform.
If connecting to multiple databases is a new requirement, then you're likely going to have to change much of your code / environment anyway, no matter whether or not you're using Django 1.0, 1.2, or Rails.
Also, for what it's worth, the OP doesn't mention just Django, it mentions Pylons, Repoze, etc. Django isn't the only Python tool in the arsenal.
Besides that, if you swap out your database abstraction layer in almost any language that I know of, you're going to have bigger problems than 'finding new plug-ins'. I can't think of a single ORM that is replaceable to another without significant retooling above the model layer. Most notably though, since you cite it as an example, losing the admin panel can hardly be considered as too big a deal, since the last time I looked at Rails, Symfony or Cake, none of them even offered that as a feature.
As for your last sentence:
I'd rather use one tool that should cover my needs than be forced to switch them because they don't fit the needs of project.
Can I safely assume that you're building everything in Java then? I mean, honestly, we pick the right tool for the job, don't we? We shouldn't be building things in Django because it's all we know, just as we shouldn't be building things in Ruby/Rails for that reason. Take a look at the challenges and figure out what makes sense.
If you don't know wha the challenges are before you start developing (they're often surprising), then just start iterating until you figure out what they are. In this case, whatever framework or development platform lets you iterate the most quickly is the winner, whatever language it happens to be written in.
How about you read that again, I think you misunderstood my point - or I was uncler, since english is not my native language. And i'm the OP in case you didn't notice that;-)
"I'm curious why you think Django lacks quality and maturity." - This is what he wrote, so i gave him main reason why django is not the tool for me.
Which are perfectly valid in my opinion. And no - with sqlalchemys design, it DOES NOT require you to change your code if you just need to use few models with other databases - the change was about 20 lines of code in my case. Which i take would be not possible at all with django 1.0.
You are right about "right tool for the job", for the kind of work I do, glue frameworks like pylons or bfg are a lot better choice - it's really way harder to corner me with some requirement that this stack can't handle.
And as for challenges, like I said - business conditions changed and I had to connect to 3 db's at once, no problems at all - if one chooses his tools very carefully in advance that can save a lot of headache later. Like I said, sorry for confusing you.
You can't predict everything that client could need/forgot to mention, so the more versatile your toolbox is the better.
That's the point, why work with alternatives if you can bear all your stack with great dynamic and fast frameworks. Endorsing pycurracy is the worst part of your argument, this framework is terrible, its codebase is far away from mature, its TOTALY regex based what makes the testing experience painful.
I love python for many reasons, but I ain't convinced yet for web and tests
How about something simple like pretending there we have django 1.0 (not 1.2 which is quite new), and let us try to connect to 3 databases at once with django's ORM?
Oh, we are screwed it can't do that. Ok, then lets try to swap out our ORM to sqlalchemy.... crap all nice plugins broke, admin panel doesn't work etc. This is why REALLY django lack quality and maturity - it's components are not versatile enough - and in my book this is a very serious issue because if my requirements are diffrent from what django makers assumed then I'm doomed. This is why people mention other (in my opinion superior, but not that easy to use), solutions to the problem. My point is - majority of people fail to notice such serious flaws like unability to use more than 1DB at once (yes i know it got finally fixed after few years). I'd rather use one tool that should cover my needs than be forced to switch them because they don't fit the needs of project.