[webkit-dev] PageCache and modularization (was Re: The Care and Feeding of WebCore Modules)

[Adam Barth]

On Mon, Mar 5, 2012 at 8:47 AM, Brady Eidson <beidson at apple.com> wrote:
> On Mar 2, 2012, at 3:10 PM, Adam Barth <abarth at webkit.org> wrote:
>> 2012/3/2 Alexey Proskuryakov <ap at webkit.org>:
>>> It seems that we have some agreement on what to modularize, but not necessarily agreement on what the modularization should encompass.
>>>
>>> For example, I think that <https://bugs.webkit.org/show_bug.cgi?id=80193> that moves some knowledge of what makes a page uncacheable out of PageCache::canCachePageContainingThisFrame makes the code less hackable, and is likely to become a disadvantage.
>>
>> I've just started investigating that issue (which is why the patch is
>> labeled "work-in-progress").  Is there a reason SQLDatabase is special
>> cased in PageCache?  Most features interact with the PageCache via
>> ActiveDOMObject::canSuspend().  In that bug, I'm investigating why
>> SQLDatabase doesn't just use the same general mechanism as other
>> features.
>
> There's no great mystery.  SQLDatabase was added before we had the ActiveDOMObject mechanism.

Makes sense.

2012/3/5 Alexey Proskuryakov <ap at webkit.org>:
> It's difficult to make something an ActiveDOMObjects without introducing subtle bugs. The automatic invocation of suspend/resume functions it provides comes at the cost of unpredictable order they are invoked in, and of very strict limitations on what these functions can do.

For better or worse, Geolocation doesn't support suspend/resume
(yet?), so this isn't much of a concern for this patch.

Thanks,
Adam