Lists Home |
Date Index |
Björn Höhrmann wrote
> * Rick Jelliffe wrote:
>>1) Tough and regular layering review (e.g. by TAG?) to refactor the spec
>>and groups as they grow. XML Schemas is under-layered and large =
> Many W3C Working Groups try very hard to minimize any
> work in response to reviews.
> It seems highly unlikely though that such additional review would come
> from an organization inside W3C, they don't have resources for that;
> for many working groups it's difficult enough to work on their own
A large, undelayered specification becomes very hard to reason about,
and there are many interdependencies, often undocumented or unclear during
the earlier days of committee work. On such large specs, problems are
either difficult to give an answer too or are so entrenched that reviews
become useless: a problem is pointed out but it is too entrenched to do
I think the issue is how to prevent Working Groups and specs which have
little chance of producing good specifications and technologies with an
open process that is responsive to reviews.
>>2) Maximum two-year membership of working groups (2 years on, 2 years
>>off). This will encourage smaller specs and encourage fresh eyes on
> There are two problems here though, you would lose much expertise
> after these two years and I'm afraid many W3C Activities have a
> rather small pool of people who could work efficiently on specific
But losing expertise is the point. At some stage, the young thrusting
achievers become the corrupt old guard, unable and unaware of other
possibilities and bound by career motives to push for the centrality of
their concerns and expertise. You want some churn, fresh blood, etc.
Keeping a two year group membership would alter the dynamic of spec
development and encourage layered, non-monlithic specs that each were more
implementable. The old guard will still be around for advise and comment.
And they can join other working groups to help inter-specification
Look at XML Schemas: big fat type derivation system which is not much
value (for complex types) and XQuery still needed to rework the derivations
for datatypes. People think "oh we need a big spec because we need to
feed back later requirements into the earlier stages" but even when they
do, it still doesn't work.