Re: New split-out drafts vs. modular design

On Jan 10, 2010, at 12:16 PM, Lachlan Hunt wrote:

> Larry Masinter wrote:
>> The problem is not that HTML5 is described in a single spec,
>> the problem is that the WhatWG design for the Web Hypertext
>> Application Platform is not modular enough. Taking a monolithic
>> design and hacking the specification into pieces doesn't help
>> address the design problem, unless the split into separate
>> specifications is used as an opportunity to remove
>> interdependencies so that each element of the design can
>> be read, understood, and implemented independently, and
>> evolved asynchronously -- on their own schedules and
>> development timelines, and likely in separate groups
>> with focused expertise.
>>
>> It may well be that there are implementation considerations
>> that cross modularity boundaries, but those should be
>> exceptions, and would best be placed in an implementor's
>> guide, not threaded ad hoc into multiple specifications.
>
> From my past experience doing QA within Opera for our implemetation  
> of  various parts of the spec, it is significantly easier to have  
> the spec define both how something is to be used and how it is to be  
> implemented in the same spec.  If these details were somehow split  
> between an authoring spec and an implementer spec as you seem to be  
> suggesting, my job would be made significantly more difficult by  
> making it harder to understand how things are supposed to work,  
> since there would inherently be a lot of cross references between  
> the 2 specs, no matter what.

I strongly agree that separating the requirements for content  
producers and content consumers makes things difficult. In the context  
of the WebKit project, not only do we have dedicated QA staff at  
various companies, but we also require every contributor to submit  
appropriate regression tests with every change. Thus, nearly all of  
our contributors (nearly 800 unique all-time contributors at time of  
writing) are put in this same QA role. When making test cases, it's  
critical to understand both the implementation and authoring  
requirements, and how they relate. Thus, splitting these into separate  
documents could significantly hurt the quality of implementations  
(much as having too much in one document can do so).

Regards,
Maciej

Received on Monday, 11 January 2010 14:37:08 UTC