Re: my take on PHP 6, part 2
| From: | Johannes Schlüter | Date: | Tue, 18 Feb 2014 17:36:36 +0000 |
| Subject: | Re: my take on PHP 6, part 2 | ||
| References: | 1 2 3 4 5 6 7 8 9 | Groups: | php.internals |
| Request: | Send a blank email to internals+get-72688@lists.php.net to get a copy of this message | ||
On Tue, 2014-02-18 at 10:04 -0600, Ralph Schindler wrote:
>
> >>
> >> You're assuming the target implementation is trivial, or even possible.
> >
> > I know extensions are needed. Still I think PHP is powerful and
> often (often != always) fast enough.
>
> You're right for sure. By and large, the argument of the extension
> vs.
> userland debate is far less technical and more social/cost-benfit
> driven than anything else. PHP itself could change some very minor
> technical details in the engine (like enabling dl() for example), that
> would completely shift the way people build and manage projects (and
> their dependencies).
>
This is *not* a minor thing. My initial mail in this sub-thread listed
the pure minimum of things to solve. dl() is useless if PHP crashes on
following requests as an extension didn't clean up while unloading or
loading fails as the operating system didn't clean up fully.
Yes, dl() right now won't directly break with some extensions on some
operating systems ... but there are too many fragile pieces for serious
production use - that's why we limited dl().
If there is somebody with knowledge about runtime linking, loading and
unloading they are very welcome to see how to make it stable and where
to limit PHP internal API. I'd love to get rid of global pecl installs
but its unfortunately not a "minor technical detail"... hrm thinking
about deployments etc. I think it is even more complicated as we'd have
to define a clean stable API else upgrading etc becomes a mess ...
johannes