Ron:

        You are correct that the library will typically use less space,
however ... This is very much dependent on the size of the CSECTS within
it.  I don't have a problem with the assertion that 32760 results in
larger TXT blocks being written, but unless you have a library that has
such large blocks, it is not a foregone conclusion that you'll save
anything.

        In doing a couple of experiments, SYS1.LINKLIB resulted in the
following:

        Blocked from 6144 to 32760 - 13% savings
        Blocked from 6144 to 23476 - 11% savings
        Blocked from 6144 to 13030 - 9% savings

Interestingly enough when COPYMOD was used to reblock it to 6144 from
6144 there was a net savings of 2%. (In all cases, the library was
copied fresh, so there was no need for compression to remove "dead"
space.)

We can surmise from these savings that we are primarily dependent on the
average size of the TXT records in the load module.  The largest
incremental savings occurred when blocking doubled to 13030, so the
primary benefit occurred for CSECTS that were between 8 - 12K in size.
When we increased the blocksize again, we picked up those few CSECTS
that were larger, and so on. 

        SYS1.LPALIB gained a savings of 33% from the reblocking (6144 to
32760).

        CAILIB only gained about 8% savings.

        But, as I stated earlier, these libraries probably contain the
largest load modules and consequently might benefit to varying degrees
by re-blocking.  However, based on the observed 2% savings using
COPYMOD, I do wonder how much of the "savings" would have been lost
through normal use of the Binder.  

        Let me re-iterate.  I don't dispute the fact that the larger
blocksizes will result in larger TXT records being written which can
result in space savings.  I guess what I'm trying to say is that I don't
find it that exciting to save 2% or 4% in a dataset.

        I'm also suggesting that despite the 32760 blocksize
recommendations, perhaps it simply isn't worth the effort in most cases
to worry about changing how libraries are distributed.  (I know I
certainly wouldn't lead a crusade to reblock all my installation load
libraries).

Anyway ... Thanks for your comments.

Adam

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Reply via email to