incubator-cassandra-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Edward Ribeiro <edward.ribe...@gmail.com>
Subject Re: Column is immutable ... no, it's not
Date Wed, 08 Jul 2009 02:11:56 GMT
Sorry, I didn't mean to offend nobody, and Cassandra is great.
Nevertheless, since I screened the code the first time, the
non-optimal, C++ like stuff just caught my attention. It's my fault,
sorry.

On Tue, Jul 7, 2009 at 10:21 PM, Jonathan Ellis<jbellis@gmail.com> wrote:
> Please give us some credit.  It's obvious that you can modify the
> contents of the byte[], but we're not trying to protect against
> malicious plugins or... something; we're only concerned with thread
> safety.

But this small fault makes the code *non* thread safe at all... If the
code cannot stand an array copy then we can just follow the first
option: remove the immutable declaration in the javadoc, because it's
*not* true. A false statement can be worse than a bug, you know.
That's what I said. Nobody is able to predict the use of this code
four years from now, who knows. But if we can take the risk then it's
ok to me.

As for the "micro-cleaning in the vaccum", I agree that they are
micro-cleaning, but they crippled through the code (this makes a big
mess?) and I believe that if you don't do this now, when Cassandra is
0.3 then we'll *never* to this when it's 1.0. Cassandra code is
cleaner and smaller than Hadoop now, but this situation can degenerate
quickly if we become permissible to such 'irrelevant' things. The
sympton is clear: it will become harder and harder to add or fix
features.

But you're right. The best strategy is picking the feature oriented
stuff. I myself got immersed in a large java project to devote any
serious attention to Cassandra. From now on I will try to avoid being
bogged down by non feature oriented stuff.

Again, I didn't want to offend nobody and Cassandra is a really
promising project. :)

Cheers,
Edward

Mime
View raw message