TEST

This must be the first record in any dataset. This contains all of the information needed to determine how to read the dataset. The information is stored in 8-bit bytes so that it is machine-independent. A reader program on any computer system should be able to determine the format of the dataset, whether or not it can read that dataset, and how to read it.

The dataset is defined to begin at the TEST record, no matter what information appears before it. It is permissible to include miscellaneous information before this record in the dataset (such as a Standard Format Data Unit label); such prepended information will not be considered to be a part of the dataset for the purposes of this standard. If the dataset is written with some sort of header bytes before the TEST record (e.g., the byte array header in the XDR format) then the dataset is defined to start with that header. The presence of these headers can be deduced from the contents of the TEST record. Thus, when trying to identify a dataset of unknown type, a program should scan until it encounters the magic number, read the next 20 bytes, and determine whether header bytes are present

The record consists of a total of 24 bytes.

name

bytes

description

MAGIC

• [1-4 ] contains the magic number: 47 f3 46 e3 in hexadecimal notation. The bytes must be specified in exactly this order. The dataset can be searched byte for byte until this sequence is found. The odds of this sequence occurring at random is 1 in 4294967296. Therefore, there is a slight chance of a false identification of a dataset being in the standard format. There is also a very slight chance that prepended information in a standard dataset will be identified as the start of the dataset.

MACHID

5

contains an identification number of the machine that generated the dataset. This field is for tracking the history of the dataset and is not needed for interpreting the dataset. The codes are as follows:

0	=	unknown (free use of this code is discouraged)
1	=	DEC VAX running VMS
2	=	Silicon Graphics running IRIX
3	=	Cray Y-MP running UNICOS
4	=	IBM mainframe running MVS
5	=	MS-DOS personal computer
6	=	Apple Macintosh running Mac OS
7	=	Sun workstation running SunOS
8	=	DEC Ultrix workstation
9	=	Hewlett-Packardrunning HP-UX
10	=	IBM running AIX
11	=	Convex
12	=	Linux
13	=	Microsoft Windows
14	=	A/UX or Mac OS X
15	=	OSF
15	=	RISC OS

NUMOBJECTS

6

is the number of data objects contained in the dataset. If zero, then there is assumed to be one data object

SPECA

7

contains information about the byte and word formats of the data. Bits are specified as follows (counting from the least significant bit):

CHARSET

(bits 0-1) indicates which character set is used for text as follows:

0	=	XDR (subsequent bits in this byte are ignored)
1	=	ASCII
2	=	EBCDIC

BSWAP

(bit 2) indicates what kind of byte-swapping is used as follows:

0	=	most significant byte is stored in the lower of the two machine word addresses
1	=	most significant byte is stored in the higher of the two machine word addresses

WSWAP

(bit 3) indicates what kind of word-swapping is used as follows:

0	=	most significant word is stored in the lower of the two machine word addresses
1	=	most significant word is stored in the higher of the two machine word addresses

RESERVE

(bits 4-7) is reserved for future expansion

RESERVED1

8

is reserved for future expansion

RECHDR

9

indicates what kind of operating system and language record headers are present. The intent here is to specify which algorithm to use in tracing through any headers present, not to specify which machine wrote the dataset. This field is completely independent of the MACHID field. A machine may be able to read and write headers from foreign machines. The codes are specified as follows:

0	=	XDR
1	=	none (pure stream file)
2	=	VMS Fortran segmented variable length
3	=	f77 Fortran
4	=	Cray COS
5	=	IBM VBS

RESERVED2

10

is reserved for future expansion

SPECB

11

contains information about array ordering and index references. Bits are specified as follows (counting from the least significant bit):

ARRORD

(bit 0) indicates the ordering of elements in multidimensional arrays:

0	=	fastest-varying index is last (most languages)
1	=	slowest-varying index is last (Fortran, IDL)

IDXSTART

(bit 1) indicates the starting value of indices:

0	=	start from 0
1	=	start from 1

RESERVE

(bits 2-7) is reserved for future expansion

SLEN

12

indicates the length in bits of a short integer

0

=

XDR

LLEN

13

indicates the length in bits of a long integer

0

=

XDR

FLEN

14

indicates the length in bits of a single precision floating-point number

0

=

XDR

DLEN

15

indicates the length in bits of a double precision floating-point number

0

=

XDR

FPFORM

16

contains information about the form of floating-point numbers. Bits are specified as follows (counting from the least significant bit):

SP

(bits 0-3) indicates the format code of a single precision floating-point number:

0	=	XDR
1	=	IEEE
2	=	VAX
3	=	IBM mainframe
4	=	Cray

DP

(bits 4-7) indicates the format code of a double precision floating-point number:

0	=	XDR
1	=	IEEE
2	=	VAX D
3	=	IBM mainframe
4	=	Cray
5	=	VAX G

RESERVED3

17

is reserved for future expansion

RESERVED4

18

is reserved for future expansion

RESERVED5

19

is reserved for future expansion

RESERVED6

20

is reserved for future expansion

RESERVED7

21

is reserved for future expansion

RESERVED8

22

is reserved for future expansion

RESERVED9

23

is reserved for future expansion

RESERVED10

24

is reserved for future expansion