Github user tillrohrmann commented on the pull request:
https://github.com/apache/incubator-flink/pull/74#issuecomment-49416463
The underlying problem is that Flink can use different serializers for
serialized objects in one stream. Thus, it is not possible to use the Kryo
serializer all the time. If that weren't the case, it would not matter whether
Kryo already read some data into its buffer belonging to the next object.
If we have a memory abstraction instead of a stream, as proposed in
[FLINK-987](https://issues.apache.org/jira/browse/FLINK-987), which allows to
seek the current pointer, we should be able to virtually write data back which
has been read from a different object. However, that would require that we can
seek backwards over memory segment borders.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
