Dear Mr. Buliabyak, thank you for the explanation of the complexity of implementation of the "baseline-shift" tag and tspan sections into inkscape. I'm ready for helping with the development of the application, but there is one big obstacle: I can't do C language.
There are many ways to help other than coding: testing, documentation, discussions...
I understand, that you - and the development team as well - are not interested in the text part of SVG.
This is wrong. We are doing a lot for text. Since we forked from Sodipodi, we added letterspacing and kerning, and recently switched to Pango for much better Unicode text support (not counting hundreds of bugfixes). All this is necessary for any further improvements of text support.
As the SVG is only suitable vector graphics format, many chemical structure drawing programs have export to the SVG. And as we need some other graphic elements (frames, boxes, splines) when we prepare chemical structures for the presentations or publications, the only way is usage of some vector graphic editor. Chemical structures are made of thin lines, the chemical bonds, and the export to the raster image degrade such lines completely. But alas, almost every such chemical structure include the subscripted text. There is no way through.
Right now you can approximate sub/superscripts by vertical kerning. For example type "H2O", position text cursor before 2 and press Alt+down a few times, then move after 2 and press Alt+up a few times. The only problem with this method is that you can't make "2" smaller than "H". See
http://inkscape.org/screenshots/gallery/inkscape-0.38-linux-kerning.png
for an example of what is possible using kerning.
I do not understand the reason for existence of two similar projects on the sourceforge, inkscape and sodipodi.
See the FAQ on our site. In short, Sodipodi was too closed and inflexible so we had to start a new project from Sodipodi codebase so we could all work freely on what we wanted to see in a vector editor.
I look into the sodipodi source code and in my opinion to extend the tspan routines is not such a big problem (if you don't need GUI for editing of tspans, but only the interpretation of it when drawing on the canvas).
Yes, the rendering part is easier to do, but what is the use if you cannot create a subscript except by manual XML editing?