Sorry, probably not ideal to just be seeing this now given KIP deadline, but in general straightforward ones where we expect things to be uncontroversial, I think its fine to kick off a vote thread. Worst case we have people re-file votes if something substantial changes in the proposal.
-Ewen On Sat, May 19, 2018 at 6:20 AM Randall Hauch <rha...@gmail.com> wrote: > Considering this KIP is straightforward, what do you think about kicking > off a vote? Or does it need more discussion time? > > Regards, > Randall > > > On May 18, 2018, at 4:30 PM, Ewen Cheslack-Postava <e...@confluent.io> > wrote: > > > > Yeah, the usefulness of short seems questionable, but int is probably a > > large enough range for some identifiers (e.g. we use an int in schema > > registry). But yeah, I don't really have a problem with having Converters > > for each of the existing serdes just for symmetry and since presumably > > somebody finds them useful for something if they exist. > > > > -Ewen > > > >> On Fri, May 18, 2018 at 11:55 AM Randall Hauch <rha...@gmail.com> > wrote: > >> > >> Thanks, Ewen. > >> > >> You make several good points, and I've updated the KIP to hopefully > address > >> your comments. I think the symmetry with the Kafka serdes is useful, so > >> I've kept all 5 converters in the KIP. > >> > >> Interestingly, perhaps the short and int converters (with the reduced > >> ranges) are not necessarily that useful for keys either. > >> > >> Regards, > >> > >> Randall > >> > >> On Thu, May 17, 2018 at 10:08 PM, Ewen Cheslack-Postava < > e...@confluent.io > >>> > >> wrote: > >> > >>> Just a couple of minor points that don't really affect the > >> implementation: > >>> > >>> * For nulls, let's just mention the underlying serializers already > >> support > >>> this. I'm actually not sure why they should/need to, but given they do, > >>> let's just defer to that implementation. > >>> * I'm not sure where Float and Double converters are actually useful. > The > >>> use cases I know for integer serdes is for keys, but floats seem like a > >> bad > >>> choice for keys. These aren't a lot of overhead to build and maintain, > >> but > >>> if we don't know use cases for the specific types, it might be silly to > >>> spend time and effort building and maintaining them. > >>> > >>> Otherwise, this seems simple and straightforward. Generally +1 on the > >>> proposal. > >>> > >>> -Ewen > >>> > >>> On Thu, May 17, 2018 at 6:04 PM Magesh Nandakumar < > mage...@confluent.io> > >>> wrote: > >>> > >>>> Thanks Randall for the KIP. I think it will be super useful and looks > >>>> pretty straightforward to me. > >>>> > >>>> Thanks > >>>> Magesh > >>>> > >>>> On Thu, May 17, 2018 at 4:15 PM, Randall Hauch <rha...@gmail.com> > >> wrote: > >>>> > >>>>> I'd like to start discussion of a very straightforward proposal for > >>>> Connect > >>>>> to add converters for the basic primitive number types: integer, > >> short, > >>>>> long, double, and float. Here is the KIP: > >>>>> > >>>>> https://cwiki.apache.org/confluence/display/KAFKA/KIP- > >>>>> 305%3A+Add+Connect+primitive+number+converters > >>>>> > >>>>> As mentioned in the KIP, I've created a pull request ( > >>>>> https://github.com/apache/kafka/pull/5034) for those looking for > >>>>> implementation details. > >>>>> > >>>>> Any feedback is appreciated. > >>>>> > >>>>> Best regards, > >>>>> > >>>>> Randall > >>>>> > >>>> > >>> > >> >
