[...] > > "Extending [the icon mechanism] right" to me means following the > > current api - like adding a new icon_size const. Of course if > > vector > > The current API is also more or less fixed to B_CMAP8. GetTrackerIcon returns a BBitmap - I was just proposing adding another enum member to return a bigger bitmap with a non CMAP8 colorspace. B_MINI_ICON and B_LARGE_ICON would still be kept for backwards compatibility, and would still be CMAP8 of course. [...] > > 2 possible directions: > > 1) Stick to the current bitmap approach, but include support > > for > > a > > larger bitmap with more colours that can be resized as needed (ala > > MacOSX). In this case a minimal API change would enable support to > > be > > added for R1. > > 2) Go for a full vector approach - this has more dependencies > > and > > would not be feasible for an R1 release, but would bring > > theoretically > > limitless flexibility for the size of icons that could be displayed > > (although I doubt how useful that really is). > > Why do you think we need decide for a direction? To me, both > directions > are pretty much the same, and will be realized with the same API. The > main point is that the system should be fast and flexible - there is > no > reason not to support both: we should try different formats and adopt > the best ones, vector and bitmap based. Simply that if the bitmap direction was decided on, it could be added without too much hassle. I like the idea of keeping the whole thing very flexible but I recognise that will be more of an API change so won't happen in R1. No worries, good things come to those who wait. Simon