NCL vs. SML

Notes on a Comparison of Intel's Network Classification Language (NCL) with SML

NCL code example: datagram header declaration

protocol ip {

	vershl	{ ip[0:1] }			/*  [position:size] in terms of bytes */
	vers	{ (vershl & 0xf0)>>4 }	/*  getting the higher order four bits and shifting them down */
	tmphl	{ (vershl & 0x0f) }		/*  getting the lower order bits */
	hlen	{ tmphl << 2 }		/*  multiplying length by 4 to get it in terms of bytes */
	totlen	{ ip[2:2] }
	ident	{ ip[4:2] }
	frags	{ ip[6:2] }
	ttl	{ ip[8:1] }
	proto	{ ip[9:1] }
	cksum	{ ip[10:2] }
	source	{ ip[12:4] }
	dest	{ ip[16:4] }

	demux {				/* this is used to specify encapsulation */
	( proto == 6 )		{ tcp at hlen }
	( proto == 17 )		{ udp at hlen }
	( default )			{ ipunknown at hlen }
	}
}

SML comparison

In SML a datatype with names for its fields can be declared using a record declaration.
It is not possible to declare the position and length of the fields in the same way as it is in NCL. The order of the fields is changed in SML so that the fields are in lexicographic order. The size of a field is determined by its type.
SML has a Word library that has 8-bit and 32-bit words but not 16-bit words.
NCL uses the keyword demux to specify encapsulation. Each item consists of a predicate and an encapsulation. The encapsulation is defined whenever the predicate is true. SML does not provide this kind of classification capabilities in the type declarations; it is not possible to have a record with a field whose type depends on the value of another field.

Equivalent SML partial code

type ipdatagram = {
	vershl 	: Word8.word,
     (*	vers 	: 
	tmphl 	:
	hlen	: *)
     (* totlen	: 16 bits
	ident	: 16 bits
	frags	: 16 bits *)
	ttl	: Word8.word,
	proto	: Word8.word,
     (*	cksum	: 16 bits *)
	source	: Word32.word,
	dest	: Word32.word,
}

Last updated: July 16 '03