In this paper we describe a GKS implementation made at ECMWF and we point out some problems encountered related to the definition of GKS 6.2 and to the necessity to adapt to the meteorogical applications.