On Wed, 2 Sep 2015 07:32:15 -0400, John Eells wrote:

>[email protected] wrote:
><snip>
>> Mind you, max blocksize is 32768, but that info is useless. AMBLIST can 
>> *probably* help you, but ...
>
>The maximum block size for z/OS is actually 32760.  (We have to reserve
>space for those pesky RDWs and BDWs.)
> 
No, the block size counts the BDW and RDWs, so the longest data portion
is 32752.

The count in a CCW allows up to 65535.  Originally, a smaller value was chosen
because the s/360 hardware has no direct support for unsigned halfword
arithmetic.  It could have been done, SMOP, but with added code at a time
when storage was precious.  And using more than 32 Kib for a buffer was
likewise unthinkable.

Why not 32767?  The best explanation I've gotten here is that the access
method requires a few overhead bytes with each buffer, and there is (was?)
a desire to obtain buffer storage page-aligned.

Bad History.  Otherwise, full track on a 3390 would be practical.

And 32768 is reserved for LRECL=X.

-- gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to