I have defended worse position in my carreer so here is a try.
Even if I lack motivation since I dont enter this line.

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6
5. Nc3 g6 6. Be3 Bg7 7. f3 O-O 8. Qd2 Nc6 9. Bc4 Bd7 10. O-O-O Rc8 11. Bb3 Ne5
12. h4 h5 13. Bg5 Rc5 14. g4 hxg4 15. f4 Nc4 16. Qe2 Qc8 17. Bxf6 Bxf6 18. Nd5
Rxd5 19. exd5 b5 20. h5 g5 21. fxg5 Bxg5+ 22. Kb1 f5 23. h6 f4 24. Rh5 Bf6 25.
Qe4 Qe8 26. Rhh1 Ne3 27. Rd3 Be5 28. Rxe3 fxe3 29. Nc6 (29. Ne6 Bxe6 30. dxe6
Qh5 {and black wins}) 29... Qf7 30. a4 Qf6 31. axb5 (31. Qxe3 Bxb2 {
looks equal to me} 32. axb5 Ba3 33. c3 Bc5 34. Qd2 g3 35. Bc2 Bf2) 31... Qg5
32. Nxe5 Qxe5 33. Qxe5 dxe5 34. d6+ Kh7 35. dxe7 Re8 36. Bd5 Bf5 37. Kc1 Rxe7 {
is a draw according to Rybka} *

sorry let me be clear about where i am with these lines i quote

after Rh5 Bf6 Qe4 Qe8 Rhh1 Ne3 Rd3 Be5 which you claim is equal,
i propose Rxe3!? fxe3 and now either Nc6 or Ne6 lines

Nc6 Qf7!? a4!? bxa4 Nxe5 dxe5 Bc4 where i would argue that blacks abysmal structure combined with the initiaive granted to white via the means of the passed h6 pawn give white the edge. alternative responses welcome

Ne6
Ne6 Bxe6 dxe6 Bf4!? c3!? maybe Rf4 is also better than Bf4 but i dont know at all it is complex. may be interesting. Rf4 Qxe3 and now Qf8! which crafty misses completly may be strong for black.
Qxe3 will run into trouble after Qf7.

defend the black side if you dare.

I am not home at the moment by qucik Fritz check suggests that after e5 Na4! is strong

g2 is analsyed. But i did not mention 23 __ Kg7, it might be playable, I will have a look at it.

Kg7 leeds to complicated endings where white is a whole bishop up but blacks passed pawns are very advanced and since both sides has Queens left the position is difficult to asses.

alternative is leeds to endings where black has two pawns for the Bishop and both sides has also a rook left. 

If white diverts from he endings there are also many forced draw lines

Rykba3  preferces white in the endings, but computers generallty undervalues advanced passed pawns

Also meeting h5 with Nb6 loses by force after e5!

In my experience of analysing Dragon games/postion, dont take any position for granted, especielly not in solits lines where white (or black)has saced material! g3 is a novelty and at first sight it showed clear advantaget to black. Then after going deeper the position is in fact the opposete.!

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6
5. Nc3 g6 6. Be3 Bg7 7. f3 O-O 8. Qd2 Nc6 9. Bc4 Bd7 10. O-O-O Rc8 11. Bb3 Ne5
12. h4 h5 13. Bg5 Rc5 14. g4 hxg4 15. f4 Nc4 16. Qe2 Qc8 17. Bxf6 Bxf6 18. h5
g3 19. hxg6 Bg4 (19... fxg6 20. Qg2 Bg4 21. Rd3 {and white gets a strong attack
}) 20. Qd3 Bxd1 (20... g2 21. gxf7+ Rxf7 22. Rh2 Bxd1 23. Rxg2+ Bg7 (23... Bg4
24. Nf5 Rh7 25. Rxg4+ Kh8 26. Rg1 b5 (26... Nxb2 27. Qg3 {and white wins fast})
27. Bxc4 Rxc4 28. Nd5 {gives white an edge}) 24. Nxd1 b5 25. f5 {and white wins
}) 21. e5 Bg7 (21... dxe5 22. Bxc4 Rxc4 23. g7 Bxg7 24. Qh7#) (21... Bxc2 22.
Bxc2 g2 23. gxf7+ Rxf7 24. Rg1 Nxe5 (24... Bxe5 25. Rxg2+ Kf8 26. fxe5 Nxe5 {
and white is better}) (24... Qg4 25. exf6 Rxf6 26. Nde2 {
and white has a slight edge}) 25. fxe5 Rxe5 26. Rxg2+ Kf8 {
and white has a slight advantage}) (21... g2 22. Rxd1 dxe5 23. gxf7+ Rxf7 24.
Ne6 exf4 (24... Rg7 25. Nxg7 Kxg7 26. Qg3+ Kf7 27. fxe5 Rxe5 28. Qxg2) 25. Qg6+
Rg7 26. Nxg7 Bxg7 27. Ne4 {and white is better}) 22. Bxc4 Rxc4 23. Nd5 1-0

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6
5. Nc3 g6 6. Be3 Bg7 7. f3 O-O 8. Qd2 Nc6 9. Bc4 Bd7 10. O-O-O Rc8 11. Bb3 Ne5
12. h4 h5 13. Bg5 Rc5 14. g4 hxg4 15. f4 Nc4 16. Qe2 Qc8 17. Bxf6 Bxf6 18. Nd5
Rxd5 19. exd5 b5 20. h5 g5 21. fxg5 Bxg5+ 22. Kb1 f5 23. h6 f4 24. h7+ (24. Rh5
Bf6 25. Qe4 Qe8 26. Rhh1 Ne3 27. Rd3 Be5 {is equal}) (24. Bxc4 Qxc4 25. h7+ (
25. Qe4 {must be better with an equal position}) 25... Kh8 26. Qxc4 bxc4 {
is an other line given by Rogozenko}) 24... Kh8 25. Ne6 Bxe6 26. dxe6 Ne3 {
and black is better according to Rogozenko} *

Sorry to clear this up, but, this is not meant to be a refutation of the entire Soltis, just an assertion that the supposedly good Topalov exchange sac is not actually as good as it seems. 
Im sure that the Soltis stands good as i play it and anyway i dont play the exchange sac here i play b5 which is dearings main recommendation. 
the main focus here should be analytical because i would like to play the exchange sacrifice because it is an incredibly deep line which my computer doesnt understand at all (claims White to be 4 points better but one advance of the pawn puts black at least equal and the points continue to fall) and i have a scholarly interest in the line anyway.

So can people please begin some actual thinking on the main idea which is after 20.Nd5 Rxd5 exd5 b5 h5 g5 fxg5 Bxg5+ Kb1 f4 and now h6, with the idea of playing Rh5.

the main lines i have come up with are 

h6 f4 Rh5 Bf6 Qe4 : f3?! Ne6
                            Qc5?! Qxf4 both seem to be good for white

h6 Kh7!? Rh5 ( i can see no better continuation)

please any suggestions#