Re: RFC: PG(allow_url_fopen) slight behaviour change
| From: | Stig S. Bakken | Date: | Tue, 16 Apr 2002 22:08:49 +0000 |
| Subject: | Re: RFC: PG(allow_url_fopen) slight behaviour change | ||
| References: | 1 | Groups: | php.dev |
| Request: | Send a blank email to php-dev+get-82865@lists.php.net to get a copy of this message | ||
I think this is fine.
- Stig
On Tue, 2002-04-16 at 19:06, Wez Furlong wrote:
> There is currently (and probably has been for quite a few releases) a
> crash bug where allow_url_fopen can not be set on a per-virtual host basis.
> That's because wrappers are only initialized on module init if
> PG(allow_url_fopen) is set, rather than request init.
>
> I plan to change it so that wrappers are *always* initialized.
>
> So that the allow_url_fopen setting is respected, I'm adding an is_url
> field to the wrapper structure. If a wrapper has this field set and
> PG(allow_url_fopen) is not set, the streams system will display a warning
> stating that URLs are not allowed.
>
> Now, the question is: was the original intention of the allow_url_fopen
> setting to prevent people from running scripts that contact the network?
> If so, it would be OK to have the zlib, bzip2 and user-space wrappers
> classed as not being a url for the purposes of this security/policy check?
>
> Heres a chart summarizing the new behaviour:
>
> wrapper allow_url_fopen=on allow_url_fopen=off
> ====================================================
> http allowed disallowed
> ftp allowed disallowed
> bzip2 allowed allowed
> zlib allowed allowed
> user allowed allowed
>
> user-space streams could potentially access the network using
> fsockopen etc., but I don't think this is an issue because the
> admin can then disable the use of that function as well.
>
> Are there any reasons against this?
>
> --Wez.
>
> --
> Wez Furlong
> The Brain Room Ltd.
>
>
> --
> PHP Development Mailing List <http://www.php.net/>
> To unsubscribe, visit: http://www.php.net/unsub.php