- From: Marcos Caceres <marcos@marcosc.com>
- Date: Tue, 12 Aug 2014 10:43:51 -0400
- To: Chris Wilson <cwilso@google.com>
- Cc: Arthur Barstow <art.barstow@gmail.com>, public-openw3c@w3.org
On August 12, 2014 at 10:32:54 AM, Chris Wilson (cwilso@google.com) wrote: > On Tue, Aug 12, 2014 at 7:17 AM, Marcos Caceres wrote: > > > On August 11, 2014 at 3:51:54 PM, Chris Wilson (cwilso@google.com) wrote: > > > > Note that I'm not EXACTLY advocating for this; I actually think > > > directly using living documents in /TR is not a good idea. > > > > Can you provide the rationale as to why? > > > > Perhaps it's my impression of what most people mean by "living document". > There are different degrees of "baked", and I've seen some things go in to > living documents that I don't think necessarily represent rough consensus; Yes, I think we need to list those explicitly because they pose a threat to having living standards on TR. When things go bad, they go really bad. For example: * srcset in HTML and the huge fight we (the RICG) had to have for 2 years to get that removed from HTML in favor of <picture> (and updated srcset). Editors adding features into a spec without consensus is clearly bad and leads to a lot of confusion/frustration. <please add others> > that makes it quite hard (particularly in a large specification) to tell > what's implemented everywhere, what's pretty solidly agreed upon, and what > is just a roughed-out idea. Agree. Though specs like WHATWG HTML have per-section stability markers - though they can be deceiving. > You could, of course, use editors' drafts here, and have the living > document always represent rough consensus (i.e. EDs are branches) - and I > certainly agree we need a faster update cycle in /TR, and > I think of specs like software in this sense. We have a fast release cycle > for Chrome - at the same time, we have different channels, and they are > different levels of done. Yes, we want > > > > > > It *IS* > > > a good idea to make sure /TR documents always POINT to the current > > > spec or effort (yes, including pointing to editor's drafts). > > > > This is what we have today, no? > > > > Absolutely not. There are some specifications that do this, it's true - > but many do not. For example, a couple of weeks ago I was trying to dig up > information about the "TV" media type. I started with CSS2. I dug through > a lot of (unlinked) specs (e.g. CSS3 media queries, HTML) before > discovering the new CSS Media Queries level 4 document, which effectively > completely replaces the bit I was interested in. None of the specs in /TR > hinted that there was superceding work going on. > > (There ARE definitely counter-examples here.) Ok, making this mandatory in the templates would be great. > > I think redefining history of specs is bad. > > > > Not sure what this means. Can you provide an example? > > > > Any time you simply rev in-place without significant versioning, you're > effectively changing history. But all specs are worked using a version control system - so you always have history? Are you saying that specs should provide a link to their commit history? (if yes, many specs do - but yes, it would be nice if this was a mandatory thing).
Received on Tuesday, 12 August 2014 14:44:20 UTC