Re: binder language, binding directories, bind by copy/reference

["Buck" <buck.calabro@xxxxxxxxxxxx>]

Rick said:

> > do you mean that by hard coding your signature, you still don't need to
> > re-bind the srvpgm to it's previous callers after you add a procedure?

and Barbara replied:
> Rick, that's indeed what I mean.  By "hard-coded signature", I mean
> using a binder source with a STRPGMEXP line like this:
>
> STRPGMEXP PGMLVL( *CURRENT ) SIGNATURE("WHATEVER")
>
> As long as you never change the signature value ("WHATEVER"), programs
> using the service program never need to be recreated.

Rick wondered:
> > If so, I agree, i'm not enamored with the *current/*prv scheme either.
> > and, if so, what is the purpose of that scheme?

Well, I thought I understood the darned thing before this thread began, but
now I'm less certain.  I use BOTH methods together.

I understand Barbara's methodology:  Use one signature forever.  Never
change it, and always add new exports at the end.  Because the sig matches
the older sig, no level check.  Because the new exports are at the end, no
problems.

For some reason I can't put into words, I find this... distasteful.  Perhaps
it's just my warped personality (you think?!)  But I *like* the
*current/*prv approach.  This lets me see which client programs are using
which version of the service program, because I hard code a signature into
every version of the binder source:

STRPGMEXP PGMLVL(*CURRENT) SIGNATURE(' 2.00 01 Jun 01')
  EXPORT SYMBOL(sockOpen)
  EXPORT SYMBOL(sockRead)
  EXPORT SYMBOL(sockWrite)
  EXPORT SYMBOL(sockClose)
  EXPORT SYMBOL(returnh_errno)
  EXPORT SYMBOL(return_errno)
  EXPORT SYMBOL(a2e)
  EXPORT SYMBOL(e2a)
ENDPGMEXP
STRPGMEXP PGMLVL(*PRV) SIGNATURE(' 1.00 11 Sep 00')
  EXPORT SYMBOL(sockOpen)
  EXPORT SYMBOL(sockRead)
  EXPORT SYMBOL(sockWrite)
  EXPORT SYMBOL(sockClose)
  EXPORT SYMBOL(returnh_errno)
  EXPORT SYMBOL(return_errno)
ENDPGMEXP

Some of my client programs were compiled after Sep 2000, and don't need the
translation services.  They have never been re-compiled, and bear the
signature of version 1.  Newer clients, since June 2001 have been
re-compiled, and bear the version 2 signature.  Now, what use this
information affords me is certainly open to debate, but it just _seems_
nicer to me.

*current/*prv without signatures seems REALLY ugly to me, because machine
generated signatures are worthless for trying to figure out which clients
are using the current service program, and which are using older versions.

Goofy tales from one programmer to you,
  --buck