Re: [Oorexx-devel] [DISCUSS] Release-to-release compatibility of the compiled image.

[Mark M. <mie...@gm...>]

On 10/13/07, Rick McGuire <obj...@gm...> wrote:

> using a more incremental approach to doing this.  Some of the
> enhancements we're already discussing (such as the ::CONSTANT changes) would
> be best implemented in ways that would cause image compatibility anyway.
>
> Key to this will be implementing some changes in 3.x that will be guaranteed
> to break the programs compiled with an older version of rexxc.  And don't
> even bother mentioning having a migration utility fix things up. ....
> Some of the
> updates I'd planned for 4.0 will make it easier to make updates without
> breaking the compiled programs, which is a good thing.

This is easy for me to say, cause I don't have a large base of rexxc
tokenized programs, but it seems that in order to upgrade to a new
version of ooRexx you have to halt the interpreter and do the upgrade.
 At that point, if you had all your programs already retokenized, they
could all be replaced during the upgrade.

It seems to me that the benefits of moving forward would outweigh the
logistical work that would need to be done to replace all the existing
tokenized programs.  So, after we hash this out somewhat amongst the
people who follow this list, we could ask the wider community and see
what objections there might be.  If everyone thought like me - there'd
be no objections!  <grin>

> So, I'm sort of proposing that we plan on the 3.3.0 release being a major
> one that will not preserve the image compatibility.  I'll try to pull in as
> many of the 4.0 "image breakers" as I can at this one release boundary,
> including the changes that make it easier to maintain release-to-release
> compatibility.

This what makes sense to me:  The break in image compatibility has to
happen at some point in order to move forward.  If we make the break
now, but also try to get all the features into 3.3.0 that would make
it easier to not break image compatibility in the future, then no one
really loses anything.  Anyone following this would have to have known
that the break was coming.  I think many people probably thought the
time frame for 4.0 was around now.  So really, the difference is a
name change rather than a substantive change.

> Ok, fire away with your opinions.  A major break like this is inevitable.
> Breaking now will make it easier to implement a lot of new features, and
> potentially make it easier to move to the 4.0 release.

I think this has a big potential of making it easier to move to 4.0.
For me that is the selling point, and I think it might be the selling
point for a lot of people.

> But if the breakage
> is a major issue, then we'll just have to do the best we can until the focus
> can get shifted to 4.0.

If the image breakage is a major issue, then we certainly need people
to speak up about it.

My plan was to try and really focus on 4.0 starting around
Thanksgiving.  But, I think I can make a bigger contribution if we
work on an incremental approach.  However, either way is fine with me.
 If having 3.3.0 break image compatibility is a show stopper for
people and they speak out about it, then I'll focus more on 4.0 work
after 3.2.0 is out the door.

--
Mark Miesfeld