SHA-1 is not a replacement for MD5, but a different algorithm for the same purpose: Hashing.
As no details about the attack are posted yet we all should be very careful not to panic.
According to Bruce Schneier, the attack still needs 269 hash operations to calculate a hash, opposed to 280 operations in an "unbroken" SHA-1.
That is still a lot of work.
If the attack should be verified, SHA-1 might be dead when it comes to digital signatures and other things. For hashed passwords... you're still on a somewhat safe side (my personal current opinion).
[edit]
I should add that the proposed attack appears to be a collision attack. That means after said 269 operations you would come up with another stream of data that results in the same SHA-1 hash.
From an educated guess I'd say that this will most likely be a random string and the attacker will not have any control over the content of the colliding string.
The attack seems to be useless for means of "decrypting" a hashed password, although you might argue that this attack might open the door to other attacks that can do this - time will tell.
[/edit]
However I for one will be watching the progress of this. First of all it has to be verified by other cryptoanalysts. Then we'll see what the real implications are.
BTW using a HMAC scheme with SHA-1 should still be perfectly safe, collisions aren't an issue here.
So: Relax, read the news, learn about HMAC.
Regards,
Sal-