>If you can build a table / dbf with variable length fields, the sum of
>lengths still add up to the length of the desired record length. You see,
>in COBOL, you do not have to have all the records in a 'set' look like each
>other -- although there is typically a match key ( first nnn bytes ).
>
>Example:
>
>Invoice No / Customer Number / Customer Name / Address1 / Address2 ...
>Invoice No / P O Number / P O Date / Total amount / Sales Tax ...
>Invoice No / Line No / Description / Qty / Price / Extended ...
>Invoice No / Line No / Description / Qty / Price / Extended ...
>Invoice No / Customer Number / Customer Name / Address1 / Address2 ...
>etc.
>
>It really depends upon the receiving application, but COBOL files are
>natively, undelimited internally, described externally in the ENVIRONMENT
>and DATA DIVISIONS of the the program -- record length, blocking factor,
>etc. Now, if the application is not really COBOL, but some 'package' which
>accepts internally delimited files, then you might get by with COPY ...
>delimited.
>
>Hope this better explains one major difference between COBOL and
>self-defining relational files.

Unless I completely misunderstood the question, which is quite possible, I thought that the problem was connected with writing data stored in a FoxPro table to an SDF file. The reference to COBOL, I thought, was an analogy, and not to be taken literally.

I am, however, aware of the above. It's just been about a dozen years (thankfully) since I worked in it. Unfortunately, there's some rumblings around here about me interpreting some COBOL for someone. :-(