Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:63612
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain adamharvey.name designates 209.85.219.42 as permitted sender)
MIME-Version: 1.0
Sender: adam@adamharvey.name
In-Reply-To: <CAB75mSVg10VURjFmOhNQk87CqVP9FOQmcSzzgfMUTEb6S7WvxQ@mail.gmail.com>
References: <CAB75mSX-cMqZPTfv2vc8fakD_pr31qPR+kO5_O=zYiVoy4xH2w@mail.gmail.com>
 <CAL+t6cF=e_55=v9bZVy7Tv=V3M4CUQLueu=b5Efa+MD0ewG_qA@mail.gmail.com> <CAB75mSVg10VURjFmOhNQk87CqVP9FOQmcSzzgfMUTEb6S7WvxQ@mail.gmail.com>
Date: Thu, 25 Oct 2012 14:19:24 +0800
Message-ID: <CADajX4ALHEzXDQQuqwjiOx5bdgUVG+4pXGAJ_Unano3z9S5ZSQ@mail.gmail.com>
To: JJ <jawed@php.net>
Cc: Sherif Ramadan <theanomaly.is@gmail.com>, internals@lists.php.net
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Subject: Re: [PHP-DEV] Changing the default value of "true" for CURLOPT_SSL_VERIFYHOST
From: aharvey@php.net (Adam Harvey)

On 25 October 2012 13:46, JJ <jawed@php.net> wrote:
> On Wed, Oct 24, 2012 at 10:34 PM, Sherif Ramadan
> <theanomaly.is@gmail.com> wrote:
>> I understand there are people out there that don't read the
>> documentation and aren't aware of the difference between
>> curl_setopt($ch, CURLOPT_SSL_VERIFYHOST, 2); and curl_setopt($ch,
>> CURLOPT_SSL_VERIFYHOST, true); but still... I don't think this is a
>> good idea either.
>
> I highly doubt code that sets CURLOPT_SSL_VERIFYHOST =3D> true meant to
> imply CURLOPT_SSL_VERIFYHOST =3D> 1...which essentially bypasses host
> verification.

They may have, even in spite of it being a bad idea, since that's how
boolean =E2=86=92 integer conversion works in PHP. I don't think we can ass=
ume
that every single person who's written that line of code didn't check
whether CURLOPT_SSL_VERIFYHOST was a boolean or integer option.

> According to libcurl, CURLOPT_SSL_VERIFYHOST =3D> 1 is "not ordinarily a
> useful setting".

I agree, but I don't think we can start arbitrarily changing well
defined type conversion behaviour for one corner case. The
CURLOPT_SSL_VERIFYHOST option has been documented as expecting integer
0, 1 or 2 since at least April 2002 (and probably quite a bit earlier
than that), complete with the meanings of each value =E2=80=94 there's only=
 so
much we can do to protect developers from themselves. (In fairness,
the wording strongly recommending using option 2 only came in last
August thanks to Ilia, but nobody should have been treating the option
as a boolean option to start with.)

Fundamentally, it's a bad API on the part of curl, but that's a
separate issue. There's nothing stopping somebody proposing a saner
API on top of libcurl (as Anthony did recently with the password
hashing API atop crypt()).

In summary, I'm against changing ext/curl here.

I do have a couple of specific comments on elements of the patch
itself in the event that the changed behaviour is wanted, but I'll
post those on GitHub, since it's probably a better UI for that sort of
granular discussion.

Adam