Alan Perry wrote:
> James Carlson wrote:
>> Alan Perry wrote:
>>> Garrett D'Amore wrote:
>>>> This is not the only criteria for self-review. Self-review cases must
>>>> also be so obvious and self explanatory that no further review is
>>>> desired or required. They should usually should not be introducing
>>>> new architecture.
>>>>
>>>> I believe this case exceed this threshold, and I would like to make
>>>> sure it is properly reviewed. Please convert this to a fast track with
>>>> a one week timer.
>>> This case does not introduce new architecture. It is making changes to
>>> an architecture introduced in PSARC/2004/779.
>>>
>>> As far as the obvious criteria, obvious to whom?
>>
>> Obvious to the reviewers -- in this case, that would be those on
>> psarc-ext, such as Garrett.
>>
>> I think his response was completely appropriate. It wasn't in any way a
>> denial of anything the project team has done; it was simply a request to
>> have a couple of days to look the materials over before declaring the
>> ARC review of the change to be complete.
>>
>> Reviews outside of the ARC are a great thing, and it's good to know that
>> the project team has sought such reviews, and that there are willing
>> people with domain expertise available to provide them. They're never a
>> substitute for open ARC review, though.
>
> Does this mean that any case that requires specific knowledge of a
> particular technology is now not eligible for self-review because
> PSARC reviewers who does not work on that technology are unlikely to
> find the case obvious?
Possibly, but not necessarily. Just because you find a case "obvious"
doesn't meant that you have covered all of the concerns. Its not fair
to assume that nobody outside of the project can provide input though,
and if there is *any* doubt then fast track is always safer than
auto-approval.
>
> Excuse me while I express some frustration here. In the past, I have
> sponsored cases with more substantial changes and have been asked why
> I was wasting people's time by submitting a fast-track and not a
> self-review.
This requires human judgment. If you aren't sure, it would never hurt
to ask a member for an opinion about the right level of review.
Established architecture would be cases where you add no new interfaces,
or only project private interfaces, and the plumbing is "obvious".
Adding new interfaces for device drivers to call and use definitely
exceeds that the threshold and IMO constitutes "new architecture", and
even if it seems obvious to the project team it needs to be given the
opportunity for people outside the project team (who may or may not be
ARC members) to provide feedback.
-- Garrett
