IBM Mainframe Forum

I just want to amplify my second bullet. When the BLKSIZE is greater than 4K, each BLKSIZE unit is recorded into multiple 4K blocks, the data blocks are not streamed. If the BLKSIZE is intermediate between a 4K multiple the result is pretty bad. As an experiment I copied SYS1.MACLIB to a PDSE without changing its DCB attributes, RECFM=FB, LRECL=80, BLKSIZE=6160. Each 6160 byte block was recorded into 2 4K data blocks in the PDSE. Obviously the final data set was much larger than the original SYS1.MACLIB. If you insist on using PDSE for source type data, use BLKSIZE=4080.

I have to take back what I said in my last post. I repeated my experiment; the SYS1.MACLIB effectively copied to a PDSE with RECFM=FB, LRECL=80, BLKSIZE=4080 required the exact same space as a PDSE with RECFM=FB, LRECL=80, BLKSIZE=6160.

I don't think PDS directory blocks get reused, so if you compile your program thirty times while you are debugging it, you will use up 30 slots, even though you still only have one member...

PDS directory blocks are reused all the time - -but space used by members is not reused until reclaimed by a compress.

Stevexff - is your assumtpion based on your testing? Easy enuf to do. If not, what is the basis of your post?