[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [equinox-dev] support for component API refactoring

Hi Jeff,

I think this is a really interesting topic and it would be great if some
kind of aspect runtime could help reducing the problem of API evolution
somehow. I thought about it since you mentioned it during out short
phone conference and these are my first thoughts about it:

My colleque Stefan Roock is writing his PhD thesis about a quite related
topic (supporting framework API evolution). He analyzed in great detail
for what kinds of API changes it is possible to adapt clients
automatically and what dangerous situations might appear. I think we
could really benefit from his work to discuss the challenges and
especially the restrictions behind this idea. Therefore I would like to
invite Stefan to join the discussion. Hope this is okay with you.

I also thought about using aspects to adapting clients to class name
changes, for example. I didn't come up with an idea how to solve this. A
problem seems to be that the current AspectJ language is not able to
replace a type reference to a different type, I think. And the current
AspectJ weaver needs the type information that are used within the class
to be woven (would be a problem if you rename a class). But Matthew
might know more about the AspectJ requirements...

I could think of some kind of specialized weaving mechanism that does
not use the AspectJ implementation. But a specialized weaving would need
a specialized aspect language or API evolution language to describe the
API changes to adapt. Could be quite interesting...

Just to let you know about my first thoughts...


Jeff McAffer wrote:
To date we have operated under a 100% binary compatibility rule with an underlying assumption that the original code itself could not be modified. Over time our API has become ladden with old methods, old package and class names, old extension points, ... The challenge is keeping the API fresh, clean and clear while not breaking existing code which, according to the assumption, cannot be changed.

The importance of API maintenance should not be underestimated. It is vital both for the smooth consumption of components as they evolve but for its effect on producer/consumer trust.

With that in mind, it may be interesting to challenge the assumption the consumer's code cannot change and instead assume that the consumer does not change the code -- allow the producer change it for them. This direction was discussed amongst a number of us at the OSGi World Congress a few weeks ago. There are several flavours but here is the basic idea.

- The producer of an API, for example, changes the name of a package/class/method from A to B
- the producer also creates an aspect definition that "co-opts" all consumers and modifies references to the old package/class/method to reference the new package/class/method.
- this can capture package, class and mehod changes
- there are corner cases involving reflection and OSGi service lookup etc that may be more challenging

Similarly, a registry munging mechanism would be required to update extension/extension point usage in the face of renamed or refactored plugins.

Is there any interest in looking at this problem (I'm thinking specifically from the aspect folks). Thisis a pretty mundane use of aspects but it would be hugely useful.



equinox-dev mailing list

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature