incubator-blur-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mark Miller <markrmil...@gmail.com>
Subject Re: Heads up - HdfsDirectory in Solr
Date Mon, 17 Jun 2013 15:27:41 GMT

On Jun 17, 2013, at 11:00 AM, Aaron McCurry <amccurry@gmail.com> wrote:

> Basically my biggest concern about moving the directory and block cache
> code that Blur utilizes is loss of control.
> 
> Aaron

Yeah, I would share the same feeling. I can easily see you guys deciding that diverging is
the better path.

I think it's worth exploring collaboration as well though.

Some of my thinking:

There are two issues I would have if I was a Blur dev:

1. Loss of commit control - you are not Lucene/Solr committers.

I think this would not be difficult to solve to some degree - though there might be some initial
pain. Essentially, I would expect that perhaps you keep using your code initially, but perhaps
supply patches to Lucene - this is your code to begin with, and you guys are likely to be
the experts - I'm sure your patches would be committed quickly and as a matter of course.
It wouldn't take much of that to make some committers I would guess. If you were satisfied
how this went, you could transition over to using the Lucene HdfsDirectory. If you were not
satisfied, you would be out a bit of time contributing to Lucene - not the worst thing in
the world ;)

2. Loss of control in general - now Lucene/Solr guys get a say.

I think this one obviously has pluses and minuses. Yes, you have to deal with the Lucene/Solr
community, but you also get their brain power and hopefully contributions. A larger community
working on making the best HdfsDirectory we can.

Personally, I don't think one path blows the other out of the water, but the idea is out there.
We don't want to just suck up some of your code without offering a collaboration path - both
to be fair to you and because of the positive affect it might have on future HdfsDirectory
dev.

 - Mark


Mime
View raw message