I think Damned-Lies isn't that bad, but only does not fit our future needs because: 1. we don't use the procedure designed for GNOME 2. we don't have that many contributors. We have l10n.enlightenment.org run a Damned-Lies is because at that time others are still frustrating, and damned-lies needs less resource so we can keep track of translation statistics.
As for Rosetta, I agree that would be a mad idea for us, :) So about Transifex and Pootle now.
The one which has attracted tons of project maintainers and translators to move onto it, either running their own instance or using the public service on transifex.net. I felt its older version sucks, but new versions are exciting.
As said in the thread, the reasons why Mozilla and Libreoffice use it are mainly two points: 1. full-featured backend 2. support for many file formats other than po.
Personally I prefer Transifex over Pootle because it is more friendly to translators to know the workflow and work on their things (I and some people I know are confused at the first time they see Pootle, but have far less problems when try to use Transifex). As for the two good points of Pootle, first we use po format ONLY, second Transifex is also full-featured.
I know there would be people prefer Pootle, please feel free tell us about your concerns so it can help us make the decision, :)
There are two parts needed to be translated in our svn now, UI translations (po/*.po) and desktop entries (*.desktop). Now we can use different tools to translate po files (we are discussing this topic!), so UI translations is okay. For desktop files we are now using the flow of "download from svn -> edit -> ask people to submit". I think we can ask developers to implement something that can extract .desktop entries to po files automatically and generate translated .desktop files using a script or any way that helps. I know there are people doing such things, but not sure if there are such people among our developers.