Surprised that no one has mentioned that monitors were broken in R16B01. http://erlang.org/pipermail/erlang-bugs/2013-July/003670.html
While I do agree with general points on both sides of the minimum Erlang requirement I think its important to note that even Basho is staying with R15B01 at the moment. I haven't heard of anything major on R16B0(2|3) but given that Basho isn't running that I wonder if they found something else there. I'm also intrigued by the reason that projects have dropped R14 support. I don't know of anything super majorly awesome in newer releases so I'd wonder if it wouldn't be possible with a bit of effort to read support to upstream projects. On Wed, Jan 22, 2014 at 3:33 PM, Russell Branca <[email protected]> wrote: > The scheduler collapse problems in R15 and R16 are widely known and not > resolved. Frankly, as developers of a database, we should strive to provide > end users with the most reliable and best experience, which in my opinion > means we should recommend R14B01. There is not a battle tested, reliable > version of Erlang that has proven to solve the scheduler collapse problems, > and until that time, I think it's unwise to remove support for R14. > > > -Russell > > > On Wed, Jan 22, 2014 at 6:20 AM, Benoit Chesneau <[email protected]>wrote: > >> Robert, >> >> I understood what you meant. >> >> Imo the best thing would be creating a check list of the things that >> prevent to go to a version greater than R14. Can you share the one you have >> inside cloudant ? It will help us to reach a consensus also later to make >> sure we can fix them in next Erlang releases. >> >> This is not that I want absolutely use the latest. If we stand on an old >> and unmaintained release then we should know exactly why and check from >> time to time if we still need to stay on this version. >> >> Thanks! >> >> - benoit >> >> >> >> >> >> On Wed, Jan 22, 2014 at 2:06 PM, Robert Samuel Newson <[email protected] >> >wrote: >> >> > >> > I could have phrased it better, so I’ll do so now; >> > >> > R14 is still widely used in production and is very stable. R15 and R16 >> > have known stability problems that affect deployments using NIF’s that >> can >> > potentially run for longer than a millisecond before returning control to >> > the scheduler. >> > >> > I am not blackmailing the project but I hope you can understand how I >> feel >> > about your suggestion to remove the ability for Cloudant to continue >> > working after we are making such a large contribution and, further, >> seeking >> > to move our active development to couchdb itself. >> > >> > B. >> > >> > On 22 Jan 2014, at 13:01, Benoit Chesneau <[email protected]> wrote: >> > >> > > On Wed, Jan 22, 2014 at 1:58 PM, Dave Cottlehuber <[email protected]> >> > wrote: >> > > >> > >> On 22 January 2014 13:23, Benoit Chesneau <[email protected]> >> wrote: >> > >>> On Wed, Jan 22, 2014 at 12:41 PM, Robert Samuel Newson >> > >>> <[email protected]>wrote: >> > >>> >> > >>>> Benoit, >> > >>>> >> > >>>> Cloudant requires R14 support, it would mean our contribution to >> > couchdb >> > >>>> becomes useless to us and we could not contribute further. >> > >>>> >> > >>> >> > >>> Are you using blackmail? Is this the position of the Cloudant >> company? >> > >> >> > >> Hi Benoit, >> > >> >> > >> Your comment reads like an ad hominem attack, and I don't think Bob's >> > >> point, nor Bob, nor Cloudant, deserved that. >> > >> >> > > >> > > My questions stand. The way it is formulated, and that's not the first >> > > time, is not that clear at all. >> > >> > >>
