[13:54] cjwatson: How much work would you estimate it would take to have the graphical d-i frontend that Debian is using for Lenny working with Ubuntu? [13:55] soren: not much, I was expecting to turn it on as part of the merge [13:55] cjwatson: Oh, fantastic. [13:56] it'll take up a bit more space of course [13:57] cjwatson: What's the current exchange rate for "a bit" into megabytes? [13:57] Just an order of magnitude guess. [13:58] a few :-) [13:58] actually it might balance out because I just realised there's a bunch of stuff duplicated between the installer seed and the installer initrd, and thus effectively appearing twice on images [13:58] so if I sort that out we'll save a few megabytes again [13:59] Excellent. [13:59] anyway that much is not a problem, I'm just curious about these conversations that have apparently been happening which seem to be wanting to go further? [13:59] I don't think anything concrete came up. [14:00] ok, nijaba was pretty vague [14:00] Yes... If anything concrete did come up, I've forgotten. [14:00] the mutterings suggested that he wanted something prettier than more or less a straight conversion of the debconf questions to gtk [14:01] my preemptive advice is that this is actually quite hard :-) [14:01] I *may* have said something like "if we find something that we *really* can't do with debconf atm..." and then finished the sentence in some hand-wavy sort of way. [14:01] we might be able to do it for a very, very small number of questions (e.g. 1) [14:01] * soren nods [14:01] basically the architecture involves writing custom frontend-specific plugins in C for specific questions that need better representations [14:02] and then detecting those in the frontend-independent code and calling them as necessary [14:03] I understand.