Hi Matt,Ubiquity is nicely developed but I agree with Ikey when it comes to its maintenance or its adoption. Its design is quite complex and its using a lot of code. I think it's developed according to the current trend in the IT industry, following what's popular these days in languages like Java. It's abstract and if not OO-oriented, at least modules-centric. Pieces fit in using template patterns and though it's considered "clean" by the industry standards, it's much harder to appropriate it than a linear script full of hard-coded strings :)
It's also actively maintained by Ubuntu, and forking it makes it hard for us to keep up with the upcoming changes. We bring slight modifications to it, but so far we never tackled important things, such as partitioning strategies.
Ikey, tell me if I'm wrong, but I don't think we attempted to port/adapt Ubiquity to LMDE/Debian. From the very beginning our plan was to come up with our own installer. It also fits a long term strategy, for Ubuntu-based editions, where we can use our own codebase for the installer and start implementing things in it without being limited upstream. Some forks work fine, and other things need to be coded from scratch. With USP we managed to appropriate the code to ourselves and maintain mintmenu successfully, with Ubiquity the decision was to code our own installer. With everything we use, we need to be fully in control. Either it fits perfectly as it is, or we fork it or create our own. Ideally, we would fork existing projects and merge changes back and forth. In practice, it's often easier to write things from scratch. We don't value code or technical design as much as ideas and UIs, and even the best developed software, if its design comes with a learning curve, can represent technical debt for us.. which is something we don't want.
Another striking example is last year's mintinstall. It was among the best solutions on the market at the time. Yet, none of its code was re-used in Ubuntu's Software Center, and it was re-written from scratch by us for Mint 9. We can see ideas taken from it and put in SC, and ideas coming from SC and put in the new mintInstall... but there's not a lot of code shared by these two projects. Sometimes the right choice is to write things from scratch. Ideas matter, and results also.. the code is just the means to do it.
Regards, Clement Lefebvre Linux Mint On 17/09/10 13:46, Matt Zimmerman wrote:
On Wed, Sep 15, 2010 at 04:02:41PM +0100, Ikey Doherty wrote:We opted for a from-scratch approach after realising that porting Ubiquity was nigh-on impossible.Out of curiosity, what sort of trouble did you have in trying to adapt Ubiquity to install LMDE?