Anthony Molinaro anthonym at alumni.caltech.edu
Mon Jan 10 18:58:04 EST 2011

On Mon, Jan 10, 2011 at 04:30:58PM -0700, David Smith wrote:
> > On Fri, Jan 07, 2011 at 03:57:45PM -0800, Anthony Molinaro wrote:
> >> some minor iolist functions).  The mainline has had some very nice recent
> >> features, the handrolled parser was replaced with a leex/yecc one, and
> >> support has been added for many protobuff features (like booleans, enums,
> >> packed format, etc).   So anyway, maybe there's a way to relink the
> >> basho/erlang_protobuffs into the mainline? (again I'm a bit of a git noob
> >> so no idea if its possible to do that).
> The "minor iolist" functions make a non-trivial improvement in
> serialization performance, fwiw.

Cool, I mostly notice that at the top-level it seemed to be a slight renaming
but I didn't quite figure out what the inner part did.

> Is the mainline ngerakines branch? I don't see anything in there
> related to leex/yecc -- am I missing something?

Well, yeah, basically ngerakines is the root, current development is on
a couple different forks of that, see


The leex/yecc work and most recent active development has been happening
on freke's fork, but I've merged his changes and the basho iolist changes
onto my fork, and am working to try to get freke to merge my changes over.

> On Mon, Jan 10, 2011 at 3:48 PM, Anthony Molinaro
> <anthonym at alumni.caltech.edu> wrote:
> >
> > I also hope you guys consider joining back with the community of developers
> > working on erlang_protobuffs so riak can be used along side other protobuff
> > based systems.
> When we last checked, there was no active development on protobuffs.
> Of course we'll be happy to work with the broader community to improve
> the library. I would note, however, that it's not clear (as a
> community outsider, I reckon) where the canonical branch lives.

Well, there's not really that much of a community unfortunately, I think
most people have abandoned protobuffs in favor of thrift, but there are
a few of us who use it and are mostly forks of ngerakines at this point.
I only tend to re-evaluate whenever I need to update software, so only
recently notice that there are 2-3 people adding features into
erlang_protobuffs mostly based on the freke fork, so I merged with
that and then after figuring out I couldn't use riak-erlang-client with
this version of erlang_protobuffs I set about to correct that (thus my
fork of riak-erlang-client and this discussion).  Since freke seems to be
most active, his fork should probably be considered mainline at this point
(although if you want a mostly drop in replacement for your fork you'll want
to try mine with the few changes in my fork of riak-erlang-client).

> Also, part of the reason we've maintained our own branch is related to
> our usage of rebar for dependencies; the ngerakines branch doesn't
> currently support this. Not sure how to resolve that one, but
> something to think about.

Please have a look at my fork and at the freke fork, I believe both
use rebar and I actually have my fork setup to be pulled in by my
fork of riak-erlang-client.

So I think it's set up correctly for that.  Please let me know if its
not and I can fix it (not too horribly familiar with rebar as I mostly
use framewerk for building erlang).


Anthony Molinaro                           <anthonym at alumni.caltech.edu>

More information about the riak-users mailing list