Size of r and s in ECDSA

Currently the XML Signature 1.1 spec says:

The output of the ECDSA algorithm consists of a pair of integers usually 
referred by the pair (r, s). The signature value consists of the base64 
encoding of the concatenation of two octet-streams that respectively 
result from the octet-encoding of the values r and s in that order. 
Integer to octet-stream conversion must be done according to the I2OSP 
operation defined in the RFC 2437 <http://www.ietf.org/rfc/rfc2437.txt> 
[ PKCS1 <#ref-PKCS1> ] specification with the |l| parameter equal to the 
size of the output of the digest function in bytes (e.g. 32 for SHA-256).


But shouldn't the length of r and s be dependent on the length of the 
key, not the length of digest function?

E.g. if you are using "ecdsa-sha256",  with P521 curve, then the length 
of r and s should not be 32, but should be 66 
(521 ; round up to multiple of 8 and get 528 bits = 66 bytes)

Pratik