(This is mostly a comment on the other contributions here, but it should also add some aspects, though from a much lower level of being informed.)
We were reminded that a predominantly community-driven project cannot be steered in a way commercial development may be. In our case specifically the main commercial competitor MS can steer - and will not aim for better software by that, but for more profit and power. They also easily can afford to hire some dozen experts for a sub-project.
What I missed was an allusion to the fact that volunteers not only will just work on subjects they like, but also should only work on topics they have sufficient knowledge of.
Surface modelling and graphing is mainly a mathematical topic on the level of applicable algorithms. Of course, also MS don’t need to develop the related math anew. It’s long known - and “sponsored” by the public system of education and the worldwide network of scientists, not paid for by MS.
However, mathematically interested people also competent in programming and willing to contribute to an open and free software project, may prefer to put their efforts into something like R (statistical) or (wx)Maxima or GnuPlot or … (even gaming) rather than into dull spreadsheet software since decades bound to bad decisions now blocking promising development. (For payment some do without asking.)
The one tiny improvement I was involved in by suggesting a different mathematical approach, but where I refused to tamper with the source code, was finally implemented by a chemist (if I remember correctly).
Thus: A campaign to improve existing functions and tools from the mathematical domain should act “think big”. It’s not about fixing a mishap here and adding a pretty feature there. It should be about incorporating mathematics into Calc not only aiming for new features, but also for means to fix issues with existing tools and functions. That might be a venture also challenging high-level contributors to existing mathematical projects. Why shouldn’t spreadsheets first be enabled to act as an interface to open free mathematical software like (e.g.) wxMaxima already containg in turn the link to GnuPlot (and other already mentioned), and later (a) tailored, reduced version(s) (subset of features/modules) of that/these softwares be directly integrated into LibO?
For a recent discussion about a related topic see: [Solved] Multple regression using LINEST (View topic) • Apache OpenOffice Community Forum . (I stepped in there with the 11th post.)