From: Brian I. <in...@tt...> - 2002-08-12 07:01:40
|
I think we have some flaws in our development process. We have been stuck in a familiar rut for many many moons. The basic problem is that we want to keep on enhancing our specification to reach some nirvana point at which we'll release the 1.0 spec. I think that this is non-productive. Here's why: A lot of the little nit-picky things we are talking about really don't make any sense unless you can also talk about exactly how they'd fit into a real implementation. But we don't have any real implementations. With apologies to Why, Showell and Ingy, we don't have libyaml. And without that, it makes it really hard for me to want to discuss the finer points of vaporware. Now we're kind of in a pickle, or should I say a chickle, or perhaps just an egg. See, the main reason we don't have libyaml is because Neil refuses to code to a moving target. And god bless him. He has better things to do. I mention this now, because tommorrow I am heading up to Vancouver for a week, and I really had hoped to get some work done with Neil. To get this libyaml thing moving again. My proposal is that we just leave the spec alone until the end of the year. That may be enough time to get a stable libyaml with fairly well fleshed out Perl, Python and Ruby wrappers. Let's drop comment keys and just leave the = key asis. We can add '/' to string, and the taguri stuff. Our egg is good enough to hatch its first chick. And it's not until we have a couple chickens running around that we can know how to make a better egg. Cheers, Brian |
From: Rolf V. <rol...@he...> - 2002-08-12 09:09:17
|
Brian Ingerson wrote: > Now we're kind of in a pickle, or should I say a chickle, or perhaps > just an egg. See, the main reason we don't have libyaml is because Neil > refuses to code to a moving target. And god bless him. He has better > things to do. That's also my case for the Java implementation. > My proposal is that we just leave the spec alone until the end of the > year. That may be enough time to get a stable libyaml with fairly well > fleshed out Perl, Python and Ruby wrappers. +1. Cheers. Rolf. |
From: Robert B. <rh...@bi...> - 2002-08-12 09:15:29
|
On Mon, Aug 12, 2002 at 12:01:31AM -0700, Brian Ingerson wrote: > My proposal is that we just leave the spec alone until the end of the > year. That may be enough time to get a stable libyaml with fairly well > fleshed out Perl, Python and Ruby wrappers. I would like to second that while my knowledge on the maturity of the spec is limited. I am just watching the threads to figure out when to seriously start looking at the spec and do some coding. As a developer I can live comfortably with a suboptimal 1.0 with a list of defects/nice-to-have/fixmes which can be addressed in a later release. \rho |
From: <sh...@zi...> - 2002-08-12 18:12:00
|
> I think we have some flaws in our development process. We have been > stuck in a familiar rut for many many moons. The basic problem is > that we want to keep on enhancing our specification to reach some > nirvana point at which we'll release the 1.0 spec. I think that this is > non-productive. Here's why: > > A lot of the little nit-picky things we are talking about really don't > make any sense unless you can also talk about exactly how they'd fit > into a real implementation. But we don't have any real implementations. > With apologies to Why, Showell and Ingy, we don't have libyaml. And > without that, it makes it really hard for me to want to discuss the > finer points of vaporware. > [...] We are down to pretty nitpicky stuff, so I would hope that folks start coding anyway, confident that they will comply on important features, and that they can reimplement the nitpicky stuff later. There is a chicken-and-egg phenomenon--you need spec stability to induce implementations, but implementations will also induce spec stability. If folks just go ahead and implement their libraries, and get people using them, then I think the debate on this mailing list will move away from nitpicking, and toward issues that really matter. The pure-language implementations are maturing nicely, but we could definitely use the speed of libyaml. Just as important, though, we need to start building the layers on top of YAML, such as YPATH, YAML-RPC, YAML tutorials, YAML cookbooks, etc. I don't really care if the debates on slashes, tags, pseudo-comment keys, etc. drag on for another year, as long as it doesn't detract from more interesting topics. -- showell |
From: Brian I. <in...@tt...> - 2002-08-12 18:53:17
|
On 12/08/02 11:12 +0000, sh...@zi... wrote: > > I don't really care if the debates on slashes, tags, pseudo-comment > keys, etc. drag on for another year, as long as it doesn't detract > from more interesting topics. Neither do I. But the fact is that important implementors like Neil and Rolf are on standby until we can have a stable spec. So let's just drop the esoteric stuff from the spec and leave them as YACs that won't be considered for addition until after some specified point in time. Like after we have stable reference implementations. I think a 6 month freeze is in order. Cheers, Brian |
From: why t. l. s. <yam...@wh...> - 2002-08-12 22:25:13
|
Okay. Just got into my mailbox. Just wanted to give my agreements to this thread. A couple points in specific: Brian Ingerson (in...@tt...) wrote: > With apologies to Why, Showell and Ingy, we don't have libyaml. And > without that, it makes it really hard for me to want to discuss the > finer points of vaporware. Yeah, I've slowed work on YAML.rb to spend time going through libyaml. As much as I've enjoyed building my parser, I have complete allegiance to libyaml now. It's a fine work and I'm anxious to perform the transplant that will make libyaml the new guts to YAML.rb. > Let's drop comment keys and just leave the = key asis. We can add '/' to > string, and the taguri stuff. Words to live by. Thanks, Brian. _why |