Is it possible in Enterprise COBOL to read a z/OS UNIX byte-stream file (just a regular UNIX file, really) in chunks of varying lengths?  By this I mean I want to read, for example, the first 10 bytes, and then based on the value of those 10 bytes I may want to read a different number of bytes.

I tried by declaring the FD with RECORDING MODE U and RECORD IS VARYING DEPENDING ON x.  But it doesn't matter what I set 'x' to (or if I set it at all), it always reads the maximum size (the size of the largest 01 level record under the FD) and returns that length in 'x'.

Thanks, Frank.

No. For a READ, x will be populated by the amount of data, as you have discovered. Prior value of x is irrelevant for READ, new value set on each successful completion.

For WRITE, the value in x will be used to control the amount of data written. x will remain unchanged during that process.

How do you establish which lengths you want for any given READ? If this is for a task rather than just interest, you're going to have to read great big chunks and then process the chunks for the lengths you need.

I'd do it with a couple of ODOs but probably 99.99% would go for reference-modification :-)

We already do it "the hard way".  I just thought it could be simplified if I could read in only the amount it took me to determine the length of the next chunk to read.  Certainly you can do this in other languages like C that are more naturally byte-stream oriented.  I guess it was too much to hope for that I could do it in COBOL.

What's hard about the hard way?

There is a penalty if you only want 10 bytes and you get 500,000.

Other languages may do it, but they don't conform to the 1985 COBOL Standard :-)

You could call a C function from our COBOL program.