SDL vs SFML for long-term project

[Hydrocity]

I've used SDL before with great success in little 2D games. I'm starting a new long-term project (2D fighting game) and am trying to decide if SDL or SFML would be better in the long run. I'll be piling network play support on top of the graphics(possibly with a different library) so I'm not sure if there would be a huge performance difference. A fighting game does not require many sprites (expect in the case of a detailed background or special move). SFML is more object-oriented and thus more attractive, its simplicity would certainly make things easier.

Have any of you have good/bad experiences with either?

[Kojack]

I've never used SDL, but I've used SFML for a few years with student projects (making games like geometry wars).
I haven't really pushed the limits of it, but the older version (the 1.x version) was a lot slower at masses of sprites than Haaf's Game Engine (a directx 8 based 2d game engine). But the code is nice. I haven't tested how much 2.x is better.

The only thing I don't like about it is that the 2d vector class has very few math operators or methods. Luckily I have my own math library, but it's still a design flaw to me. Otherwise it's a nice library.

[Hydrocity]

Thanks for your opinion, Kojack. I'll start with SFML and try it for a while. I had never heard of HGE, maybe I'll have to git that a try, too.

[Eventide]

Why not Allegro? It's fast as SDL and much easier to code.

[Klaim]

I thought that Allegro was not maintained anymore but it looks like it is.