Wes McKinney

I just wanted to echo my comment from the mailing list that I think it would be best if we held a code donation vote and IP clearance for this...

We've allowed 64-bit Arrow lengths for almost a year. Keeping things under INT32_MAX is a guideline, though, for better compatibility with Java. Could you let me know what problem you're...

> I'm not entirely clear what joining the arrow project means It means building a Julia developer community in the Apache project (https://www.apache.org/foundation/how-it-works.html). It's about community and governance more than...

I wouldn't let issues related to the structure of the codebase(s) get in the way. If we need to create you a separate `apache/arrow-julia` repo on an interim basis, we...

OK, the main thing I will ask is that you keep track of whose IP is involved. There is code living in multiple repos from what I can tell, and...

@ExpandingMan Feather is limited to < 2 GB of total string data per column. This could be fixed by changing the Feather format to use the Arrow streaming/file messaging formats...

I recall a number of issues writing data frames over 4GB with the old `feather` package, there were some other issues in the past, and there may be some existing...

Whatever decision is reached here, could we incorporate this into the related Arrow patch https://github.com/apache/arrow/pull/920 so that we can round-trip to Arrow memory without any data loss?

@julianhyde my thinking in my patch https://github.com/apache/arrow/pull/920/files was to add a time resolution to the DAY_TIME interval type to address resolutions from seconds down to nanoseconds. This matches up with...

Understood. Should it become necessary in the future, we can augment the metadata to accommodate additional precision. I think 8 bytes will be adequate for the vast majority of use...