[Dataflow] [Review] Dataflow review has ended
The number of reviews for the dataflow library was very low (one, positive). Rather than extending the review period in hopes for more reviews, the review period of the data-flow library is now over, and at this time the library is not accepted as a Boost library. The discussions indicated that there's some confusion regarding the library's role; this concern needs to be addressed prior to moving forward with the library. Stjepan is content of the feedback received during the review, and will use it to clarify the library's documentation and further develop the library. He intends to bring the library for review again when the concerns raised have been addressed. Even though the review period is over, I encourage everyone to submit their comments, half-written reviews, reviews that you were planning to write, etc. as those will help Stjepan in his efforts in improving the library. Thank you for everyone for participating in the review, and thank you Stjepan for your work for Boost! Jaakko Järvi Review Manager
On Thu, Sep 11, 2008 at 9:22 AM, Jaakko Järvi <jarvi@cs.tamu.edu> wrote:
Thank you for everyone for participating in the review, and thank you Stjepan for your work for Boost!
Indeed - thanks to everyone for a really great discussion, and to Jaakko Järvi for managing this review. As a result, I now have a much better picture of where the library and the documentation need to go. For those that have expressed interest in using the library, I would appreciate any further comments you might have throughout your use. I have started a rewrite of the library (in sandbox/dataflow-rewrite), into which I am going to slowly move and solidify parts of the library, incorporating the the feedback as best as I can. When the Dataflow library is ready as a whole, I will plan to resubmit it for review. Kind regards, Stjepan
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
The number of reviews for the dataflow library was very low (one, positive).
This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?) Gennadiy
on Thu Sep 11 2008, Gennadiy Rozental <rogeeff-AT-gmail.com> wrote:
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
The number of reviews for the dataflow library was very low (one, positive).
This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?)
This seems like a good idea. There is supposed to be a "gauge interest" phase to the submission process. Is that failing to work, or are we ignoring its results, or are people claiming interest and then not submitting reviews? -- Dave Abrahams BoostPro Computing http://www.boostpro.com
On Thursday 11 September 2008 1:56:57 pm David Abrahams wrote:
on Thu Sep 11 2008, Gennadiy Rozental <rogeeff-AT-gmail.com> wrote:
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
The number of reviews for the dataflow library was very low (one, positive).
This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?)
This seems like a good idea. There is supposed to be a "gauge interest" phase to the submission process. Is that failing to work, or are we ignoring its results, or are people claiming interest and then not submitting reviews?
The pre-registration would also help review submitters schedule out time and coordinate amongst themselves an appropriate deadline that works for those interested. It's easy to indicate interest but equally easy to miss the review deadline when something regarding the day job comes up and chews up the week the review is going on. Chris
David Abrahams <dave@boostpro.com> writes:
on Thu Sep 11 2008, Gennadiy Rozental <rogeeff-AT-gmail.com> wrote:
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
The number of reviews for the dataflow library was very low (one, positive).
This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?)
This seems like a good idea. There is supposed to be a "gauge interest" phase to the submission process. Is that failing to work, or are we ignoring its results, or are people claiming interest and then not submitting reviews?
In academic conferences, reviews are conducted by program committees whose members have agreed to the task beforehand. It is very typical that almost all the reviews are submitted very close to the deadline of the reviews, often slightly after, so people clearly struggle finding the time for doing the reviews. However, at least according to my experience, the reviews eventually get written. There is no formal commitment to Boost reviews, and thus in a struggle to find time for a Boost review, I suspect Boost easily loses. Establishing a "library review committee" by some means before a review, and announcing it when the review begins might create a stronger commitment. Best, Jaakko
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
Establishing a "library review committee" by some means before a review, and announcing it when the review begins might create a stronger commitment.
One approach can be based on these ideas: Instead of dedicated library review period we can have unlimited review period. Library review should start at some point (more in this later) by review manager announcing library for review. Within some predefined period (1-2 weeks) at least N (5?) people should confirm upcoming review submission. If threshold is not meat library immediately rejected, though can be resubmitted once again later (as usual). If threshold is meat these people form library review committee. After that review can be submitted. There is no predefined review period (maybe only some lower bound). Until all people in review committee submit their review it will continue. It's responsibility of review manager to keep track of that and bug them if necessary. Everyone else are welcome to post reviews as well. The maybe some provision to avoid indefinite wait, but it should be in a form of exception clause. Under this conditions we can have multiple reviews running the same time. Boost site should announce all libraries currently under review. Review wizard will throttle the load and order libraries review in an order of submission and review author/manager availability. Review wizard might have an authority to stop any review (if it takes too long for a example) Gennadiy
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
The number of reviews for the dataflow library was very low (one, positive).
This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?)
This seems like a good idea. There is supposed to be a "gauge interest" phase to the submission process. Is that failing to work, or are we ignoring its results, or are people claiming interest and then not submitting reviews?
In general terms, given the limited resources of the volunteers for reviews, I agree that more thought could be given to the gauging of interest phase. Perhaps prior to review, an online 'vote' could be hosted on the web site to gauge general interest (new votes every release cycle to stop it becoming stale)? I do think boost could also offer a distinct means to support 'developing libraries' which perhaps allow developers a mouthpiece to announce and obtain feedback for ongoing library development either as a preliminary stage to a boost review or as an adjunct service building on the interest and availability of boost's membership. The sandbox and developer lists offer this to some extent in an ad-hoc manner but perhaps an explicit sandbox 'mini-review' checklist and early adopter's comments managed by the author as supporting evidence as part of a review request? (e.g. five positive early adopter mini reviews, plus popular vote) My perception is that currently a few motivated individuals are sufficient to get a submission in the queue and with a review slot. This may not necessarily serve the best interests of boost and the wider C++ community. I'd also be interested to understand whether boost is now rejecting more libraries after review and what the reasons have been for doing so. Is there a pattern? Perhaps potential libraries should have a tougher job making it into the review queue with regard to clear statements of purpose, wide interest, pre-existing maturity of design and ideally some maturity of implementation demonstrated at least in part by those mini-review checklist criteria.
-----Original Message----- From: boost-bounces@lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of David Abrahams Sent: 11 September 2008 19:57 To: boost@lists.boost.org Subject: Re: [boost] [Dataflow] [Review] Dataflow review has ended
on Thu Sep 11 2008, Gennadiy Rozental <rogeeff-AT-gmail.com> wrote:
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
The number of reviews for the dataflow library was very low
(one, positive).
This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?)
This seems like a good idea. There is supposed to be a "gauge interest" phase to the submission process. Is that failing to work, or are we ignoring its results, or are people claiming interest and then not submitting reviews?
Is the problem that there are far more people who are interested in *using* the library than feel confident in reviewing it? I feel that the review process intimidates the many with only a slight knowledge of the subject, but whose collection views are probably as useful (but complementary) - as one or two real experts. If we have one expert *and* a dozen people who are comfortable using the library - and using the documentation (but don't necessarily understand how it works) - that seems a good enough for acceptance to me. Paul --- Paul A Bristow Prizet Farmhouse, Kendal, Cumbria UK LA8 8AB +44 1539561830 & SMS, Mobile +44 7714 330204 & SMS pbristow@hetp.u-net.com
David Abrahams wrote:
on Thu Sep 11 2008, Gennadiy Rozental <rogeeff-AT-gmail.com> wrote:
Jaakko Järvi <jarvi <at> cs.tamu.edu> writes:
The number of reviews for the dataflow library was very low (one, positive). This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?)
This seems like a good idea. There is supposed to be a "gauge interest" phase to the submission process. Is that failing to work, or are we ignoring its results, or are people claiming interest and then not submitting reviews?
I think, one significant reason for such a situation is that the delay between such a query of interest (and submitting the library review) and the actual review is quite big. For example, the recently reviewed Boost.FSM was in the queue for more than a year, IIRC. Although there were reasons for the delay in this particular case, it's not a sole example - many libraries don't even have review managers for months/years. So when the review happens, many potential reviewers that were excited about the library while the interest poll, now either are gone or found/developed another solution for their needs. I have no concrete suggestion for improvement on this, except maybe to limit the period of waiting in the review queue. OTOH, I'm not very happy with the suggestion as a potential library submitter, since my library could be simply bounced off from the queue for reasons I have no control on.
Andrey Semashev wrote:
I think, one significant reason for such a situation is that the delay between such a query of interest (and submitting the library review) and the actual review is quite big. For example, the recently reviewed Boost.FSM was in the queue for more than a year, IIRC. Although there were reasons for the delay in this particular case, it's not a sole example - many libraries don't even have review managers for months/years. So when the review happens, many potential reviewers that were excited about the library while the interest poll, now either are gone or found/developed another solution for their needs.
I have no concrete suggestion for improvement on this,
Maybe Boost should have a review Stack instead of a review Queue. :)
Hi,
The number of reviews for the dataflow library was very low (one, positive).
This seems to be general theme nowdays. Maybe we can have some kind of pre-registration for people planing to submit the review and reschedule it if this number is less than some threshold (5?)
It would be my first review. Unfortunately I passed the date. For further handling I propose to add the review schedule on a public calendar, e.g. google calendar. This happens to the release schedule already (http://www.boost.org/community/index.html) which is linked to my personal one (Thunderbird + Lighning). To get quick results a multiple choice web page could help. Regards, Olaf PS: As you can see, I'm interested in.
participants (11)
-
Andrey Semashev -
Chris Knight -
David Abrahams -
Gennadiy Rozental -
Gennaidy Rozental
-
jarvi＠cs.tamu.edu -
Michael Marcin -
Olaf Peter -
Paul A Bristow -
Paul Baxter -
Stjepan Rajko