Protocol Augmentation

[Ed Kohlwey <ek...@gmail.com>]

Hi,
I'm working on updating Gora to use the newest Avro release-targeting
1.7.0. Gora uses a hacked serialization for Map/Reduce to augment the
Avro protocol with dirty bits, so that only data that has been
modified gets written back to the data store.

The current Avro API no longer supports this hack.

My first question is weather or not there's a "good" way to do this in
Avro. One idea is to require all Gora Avro records to extend a base
record class or to have a special field called dirty or something
similar. Obviously, this is sort of obnoxious as it requires people
introduce information into their schema that isn't really part of the
domain that the record represents.

In general, it would be nice if Avro had support for some sort of
"transient" record-level metadata or something. Some other use cases
for such a thing might include attaching Accumulo-style visibility
ACL's.

Or is this something that Avro contemplates should happen in codecs?

Any thoughts are appreciated.

-Ed Kohlwey