Speaking for myself now, and not necessarily for the TAG:
I agree with those who say or imply that it's late for making incompatible
changes to the Web Storage specification. I'm less clear that's the case
for appcache, given comments about its many problems at the workshop last
week, but just for purposes of this email let's assume that it too might be
too widely deployed to change wholesale.
I also think the TAG is right to ask that the relationship between the two
be considered more carefully than, as far as I know it has been. There are
many dimensions in which one could imagine innovating to improve the
synergies without necessarily disrupting existing deployments. To pick one
example signaled in the TAG's email, I would think that one could innovate
with application management (e.g. install/uninstall) and space management
(query space used by application, set quotas, etc.) that could be done in
ways that are compatible with the existing specs. Maybe or maybe not it
would be beneficial to integrate them further, e.g. by providing a standard
means for storing app-cached resources in Web Storage or otherwise
integrating what are now disparate clumps of application data.
Whether these will prove to be good things to do is TBD, but I agree with
the rest of the TAG that doing the work to find out is important. I also
think there are many potentially useful changes that could be made without
inappropriate disruption to early deployments.
FWIW: I would rather not debate here the difficult balance, in general,
between deploying early enough to get real user feedback before specs are
frozen, and then finding that you can't actually change the specs based on
that experience, because deployment is too widespread. That's a very
important debate, but for this thread, I would rather just concentrate on
seeing what's practical and beneficial with respect to application storage
in particular. I think there are still some things worth looking at.
Noah
On 11/15/2011 9:41 PM, Bjoern Hoehrmann wrote:
* Tab Atkins Jr. wrote:
On Tue, Nov 15, 2011 at 5:28 PM, Glenn Adams<gl...@skynav.com> wrote:
Perhaps. But widely implemented does not necessarily imply widely used. In
any case, support for or use of a feature of a WD or CR does not imply it
must be present in REC.
Use of a feature does, in fact, imply that, unless there are *very*
good reasons why not. Specs and implementations advance together, and
both constrain the other.
Well, they "advance" from Working Draft to Working Draft and then it's
too late to make changes before there is a "Call for Implementations" as
the implementations have already been shipping for years. The Last Call
is meant to avoid that, providing an opportunity to build a consensus
even with people and organizations that cannot follow the day-to-day
working group and implementation progress to prioritize their reviews.
Which the Last Calls relevant to this thread obviously do not provide.