Experimental Physics and
Industrial Control System

"J. Lewis Muir" <[email protected]> · Fri, 25 May 2012 14:19:00 -0500

On 5/25/12 1:18 PM, Mark Rivers wrote:
> Lewis,
> 
> I think your point and Dirk's are rather different.  You are
> saying "don't support old releases of modules you are
> responsible for".  Fine.  But Dirk is saying "do write your
> module X so it supports old versions of module Y that someone
> else is responsible for".  In the case of streamDevice that
> means old versions of base and asyn.
> 
> In the case of more complex modules like areaDetector, it
> would potentially mean I should support old versions of base,
> asyn, calc, busy, sscan, and autosave.
> 
> I just don't think that's realistic or a good use of my
> limited time.

Hi, Mark.

OK, maybe I misunderstood what Dirk was saying.  I agree that it
can be very time consuming to do Dirk's approach, and I agree
that, with limited time, it might not be the best choice, but
I'm just saying what Dirk did is nice because StreamDevice is
able to work with older versions of libraries which means I
could potentially upgrade just StreamDevice and nothing else.

What I thought Dirk was talking about was the concept of binary
compatibility and trying not to break that for a reasonable
amount of time.  There's a write-up about that at

http://apr.apache.org/versioning.html

.  I like this and think it would make dealing with all the
EPICS related libraries out there a lot easier because I could
very quickly tell what versions of libraries I could expect to
upgrade and still work correctly with one another.  And it would
require developers to be aware of binary compatibility since
they'd have to increment the right version number part.  Now I
just need to convince everyone to adopt this versioning scheme. :-)

Lewis

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System