Pierre On Aug 21, 2015 22:01, "Pierre Joye" <pierre....@gmail.com> wrote: > > On Sat, Aug 22, 2015 at 8:43 AM, Anthony Ferrara <ircmax...@gmail.com> wrote: > > Anatol, > > > > On Aug 21, 2015 8:10 PM, "Anatol Belski" <anatol....@belski.net> wrote: > >> > >> Hi, > >> > >> > -----Original Message----- > >> > From: Anthony Ferrara [mailto:ircmax...@gmail.com] > >> > Sent: Friday, August 21, 2015 3:37 PM > >> > To: Scott Arciszewski <sc...@paragonie.com> > >> > Cc: Pierre Joye <pierre....@gmail.com>; Trevor Suarez > >> > <ric...@gmail.com>; > >> > Niklas Keller <m...@kelunik.com>; PHP Internals < internals@lists.php.net> > >> > Subject: Re: [PHP-DEV] Recap - Core functions throwing exceptions in > >> > PHP7 > >> > > >> > On Fri, Aug 21, 2015 at 6:14 AM, Scott Arciszewski < sc...@paragonie.com> > >> > wrote: > >> > > On Fri, Aug 21, 2015 at 3:52 AM, Pierre Joye <pierre....@gmail.com> > >> > > wrote: > >> > >> On Fri, Aug 21, 2015 at 9:38 AM, Scott Arciszewski > >> > >> <sc...@paragonie.com> > >> > wrote: > >> > >>> On Wed, Aug 19, 2015 at 11:36 AM, Trevor Suarez <ric...@gmail.com > > >> > wrote: > >> > >>>> Ah, I didn't realize this thread existed. I had just commented on > >> > >>>> the old one, but the point still stands: > >> > >>>> > >> > >>>> PHP 7.0 RC1 was just tagged. > >> > >>>> Shouldn't this be a relatively high priority to fix/decide so we > >> > >>>> don't end up with behavior that can't be fixed until PHP 8.0? > >> > >>>> > >> > >>>> On Mon, Aug 10, 2015 at 6:54 PM Niklas Keller <m...@kelunik.com> > >> > >>>> wrote: > >> > >>>>> > >> > >>>>> > > >> > >>>>> > Okay, great, we have people on both sides on this discussion. I > >> > >>>>> > hope nobody minds if I sit this part out. > >> > >>>>> > > >> > >>>>> > What specifics need to be discussed? Should somebody set up a > >> > >>>>> > poll? (I don't know how to do that.) > >> > >>>>> > >> > >>>>> > >> > >>>>> You can find information on how to setup a poll in step 6 here: > >> > >>>>> https://wiki.php.net/rfc/howto > >> > >>>>> > >> > >>>>> Regards, Niklas > >> > >>> > >> > >>> I agree that this should be a relatively high priority. I'm not sure > >> > >>> what the next steps would be. (Aside: I still have a PR I need to > >> > >>> write that I've been holding off on until the fate of PHP 7's CSPRNG > >> > >>> feature is determined.) > >> > >>> > >> > >>> Can we reach some sort of consensus on throw new Exception vs throw > >> > new Error? > >> > >> > >> > >> I think the best would be a RFC, not only for the decision itself but > >> > >> also to have a clear view about what will be changed or affected. > >> > >> > >> > >> Cheers, > >> > >> -- > >> > >> Pierre > >> > >> > >> > >> @pierrejoye | http://www.libgd.org > >> > > > >> > > Fine, let's do this: > >> > > > >> > > 1. Violate the feature freeze for this exceptional decision. > >> > > 2. One of the folks in the camp that WANTS an RFC and a drawn out > >> > > formal decision-making process opens it with a poll. > >> > > 3. Give me voting karma. > >> > > > >> > > Let's NOT make the CSPRNG feature fail open. That is an absolutely > >> > > terrible idea. > >> > > >> > My proposal/stance: > >> > > >> > Let's make random_* throw an Exception if it cannot connect to a random > >> > source. And let's have it throw an TypeError if ZPP fails, or Error if > >> > min >= max. > >> > > >> > The first two are consistent with existing exceptions. > >> > > >> > The third (Error if min>max) is where the contention lies. I'm > >> > suggesting Error as > >> > it's consistent with parameter errors in the sense that the type may be > >> > correct, > >> > but the value isn't (hence it's the same kind of error as a parameter > >> > error, just a > >> > different sub-classification. > >> > > >> > MHO is this is too important of a distinction to simply gloss over. > >> > Having it return false (or null) will be a problem, as nobody will > >> > perform the error > >> > checks. And returning $x where `$x == 0` in a security context could be > >> > incredibly > >> > bad. As such, I think the security implications here outweigh nearly all > >> > of the > >> > other concerns about consistency and convention. > >> > > >> > That's my opinion. I'll be happy to make the changes if a RM gives me > >> > the green > >> > light to do so. > >> > > >> The change being proposed was discussed once more in the RM circle and is > >> being seen as inappropriate. > >> > >> The CSPRNG RFC and the implementation was voted. The change being proposed > >> amends the paradigm of the current language behavior. Currently no effort > >> has been done do discuss and work out the paradigm change. > >> > >> By today's terms, there are other functions which could require throwing > >> instead of returning false for security reasons. Security being over BC was > >> and is even in the patch versions, however how it is handled is related to > >> the hard and deeply internal cases like memory corruption, etc. Having a > >> decision that a return value is something security related has impact to the > >> existing behavior. Having different technical requirements to the congeneric > >> cases on the language level brings inconsistency. Producing inconsistent > >> behaviors by one case, without any evaluation and clear course for other > >> cases, without respecting the votes and code freeze is alarming. > >> > >> The current timeline doesn't allow for a proper solution of this topic in > >> 7.0. The RMs recommendation is that everyone expressing a strong support in > >> this thread for the behavior change either for core functions in general or > >> particularly in the security context stands up for a proper solution in 7.1. > >> If no one believes that a proper solution can exist in 7.1, then an > >> inconsistency shouldn't exist in 7.0, except the community wants it to be so > >> which brings it back to an RFC. With respect to everyone who voted on the > >> original implementation of CSPRNG RFC and everyone else regarding the topic > >> "throwing in the core functions" it should be accepted in the usual ways > >> that are foreseen. > > > > Thank you for sharing your thoughts and being transparent. > > > > There is one tiny thing I would point out though (which likely makes no > > difference). When the random rfc was voted on, engine exceptions was not > > accepted. It was a conscious decision by the contributors to not have the > > function throw because nothing throws in core. That changed with the later > > rfc. Hence why this was reopened. > > > > The discussion has been biked shedded to death. From before beta1. And > > unfortunately it looks like it has just been bike shedded out of contention > > for 7.0, which is sad on many levels. > > > > But this is where we are today. While I think it is less than optimal, so be > > it. > > I do think as well it is better to solve this question for 7.0. It is > a kind of big thing even the code changes may be small. Dealing with > that for 7.1 and 7.0 will most likely be painful. > > However we have chosen to have a short timeline to release 7.0. We > knew the risks of having such issues to solve. I personally would not > mind too much to have a RFC for this case as long as it includes an > option to slightly delay 7.0 if necessary.
If that's what it will take I will happily draft one tomorrow morning. But if the RMs are against it, I will respect that as well. Hence the dilemma. Anthony
