Re: Max length of a VARYING field

Joe Pluta wrote:

...
One option is to have ALL fields have a 48-bit index, and then we wouldn't
have to specify the index size -- that was the "adds overhead for everybody"
option. Or we can go with the extra keyword on the VARYING clause. That's
a different kind of overhead: it adds an unnecessary keyword. Talk about
making the language harder to teach! Just try explaining why you would
specify a *MAX4GB value for a field with a defined maximum length of 2KB.

You are right that it's a bit loony to say the field has a maximum length of 10, and that it also has a maximum length of *MAX4GB. But that just shows that the *MAX4GB form of specifying the size doesn't really work. Specifying prefix size as 2 or 4 doesn't say anything about the maximum length; it only says how big the storage is for the prefix.

If the prefix size changed from one release to the next, there would be chaos even if all RPG programs were recompiled. There are more ways to specify a "variable with a 2-byte prefix indicating the length" than just RPG's VARYING type. In C, you have to use a struct with a short-unsigned length; COBOL too. In a command PARM, you say VARYING(*YES:*INT2). RPG data structures may have a dependency on the size of the varying prefix. Etc, etc. There's no way IBM would make a change that would cause programmers to have locate and recode all the places that were interfacing with varying fields.

(Aside: where did the 48-bit prefix come from? That's 6 bytes; did I miss something?)