Coming back in the forum after long and far holidays I give here my "thoughts" ...  ...

The "problem" is Black has no good variation except 4...dxe4 5.Nxe4 which is transposition to Rubinstein where Black has already moved one time his f8 bishop (hence the c5 "freeing" move will lose tempo if Black want to recover pawn with bishop after dxc5...and trying to play Nd7 is no better 5...Nd7 6.Nf3 Ngf6 7.Nxf6 Bxf6 8.Be3!?...one idea is that c3 pawn move doesn't gives Black the opportunity to play Qa5+ and Qxc5) 

After 4...Nf6 5.e5 Nfd7 White has to choose between 6.Ngf3 for a Korchnoi gambit to follow and 6.Bd3 where 6...c5 is not as good as in the main line of Tarrasch (without Be7). See for example: Suchting-Holland,Hanovre 1902  (ECG = Encyclopedy of Chess Games)

After 4...c5 I will advice for White to play 5.e5 (the same position is reached via the move order 4.e5 c5 5.c3!?)
and now

A) 5...Nc6 6.Bd3 cxd4 7.cxd4 Qb6 (of course Nxd4 is not possible because of Qg4 attacking Nd4 and g7) 8.Ndf3

B) 5...cxd4 6.cxd4 Qb6/Nc6/Nh6 7.Ndf3

C) 5...Qb6 6.Qg4

D) 5...Nh6 6.Ndf3 


I add one excellent but old book written by Suetin about French gives c5?! ... but modern/authors like Psakhis/Watson prefers the "complex" assessment.

I propose (advance) variations above for White because I feel deviating from the isolated pawn game with Nf6-exd5-Nxd5 is good idea for Black.